| kota_ | ho: and good morning: ) | 00:05 |
|---|---|---|
| ho | kota_: good morning! | 00:07 |
| clayg | peluse: bah, so I have the fix in for the suffix delta's but the probe test is still failing because of some weird interation with ssync and the object server - but i'm not entirely sure it's not related to acoles_away refactoring of of the get on disk files stuff | 00:08 |
| *** vinsh has quit IRC | 00:27 | |
| *** ho has quit IRC | 00:31 | |
| *** ho has joined #openstack-swift | 00:32 | |
| *** gyee has quit IRC | 00:35 | |
| torgomatic | clayg: heh, check this diff out https://gist.github.com/smerritt/159c3b4c866a43597728 | 00:38 |
| torgomatic | all tests pass | 00:38 |
| clayg | torgomatic: I think acoles_away already said that | 00:39 |
| torgomatic | clayg: ah, I must have missed it then | 00:39 |
| notmyname | home | 00:39 |
| * notmyname goes to brew coffee and get cracking | 00:39 | |
| clayg | oh goodness - object server's can't even cook up a hashes.pkl of a hashdir is missing a .durable because of the on-disk file search algorithm contract :'( | 00:44 |
| *** ozialien has joined #openstack-swift | 00:44 | |
| clayg | i wonder if the same is true for replication with a dangling .meta | 00:45 |
| clayg | no, it seems to think that's fine | 00:47 |
| clayg | :'( | 00:47 |
| *** pberis has joined #openstack-swift | 00:47 | |
| clayg | fml | 00:47 |
| notmyname | clayg: where do you need me to start looking? | 00:48 |
| *** Guest___ has joined #openstack-swift | 00:48 | |
| clayg | i don't know I have a ton of diffs I've eaten from acoles and cschwede i haven't pushed yet - I expect to have another one from torgomatic at some point | 00:49 |
| notmyname | ok :-) | 00:49 |
| clayg | notmyname: I need something that changes/unifies the x-backend-ssync-frag-index header with whatever we're going to do there | 00:50 |
| clayg | i probably need to write a failing unittest that puts one of these "missing a .durable" directories in a tmp dir and runs it through ECDiskFileManager's get_hashes | 00:50 |
| clayg | then decide how i'm going to fix it :'( | 00:50 |
| clayg | maybe it's as simple as telling cleanup_ondisk_files to verify=False | 00:52 |
| notmyname | clayg: I'm definitely coming into the middle of that conversation. is there a place to start looking or do you want to have a quick phone chat? | 00:53 |
| * clayg wishes acoles_away didn't have to sleep | 00:53 | |
| *** Guest___ has quit IRC | 00:53 | |
| clayg | it normally works best when acoles_away just tells me what to type | 00:53 |
| notmyname | heh | 00:53 |
| clayg | notmyname: I really would appreciate a diff to the reconsttcutor patch that fixes the ssync-fragment-index header | 00:54 |
| clayg | I'm not sure what the header is supposed to be called - or what would pass as "fixes" | 00:54 |
| clayg | but I'm not working on that right now - and acoles says the current name makes him sad | 00:54 |
| notmyname | ok | 00:54 |
| notmyname | yeah, I see the comment in gerrit on it now | 00:54 |
| *** Guest___ has joined #openstack-swift | 00:55 | |
| notmyname | ok, that's where I'll start | 00:55 |
| clayg | THANKS! | 00:55 |
| notmyname | is mattoliverau on yet? | 00:55 |
| notmyname | I saw kota_ say hi | 00:56 |
| clayg | mattoliverau: yeah! how come YOU'RE not telling me what to type! | 00:56 |
| *** jrichli has joined #openstack-swift | 00:57 | |
| clayg | ok, so that part was easy -> https://gist.github.com/clayg/b75428b75ada1e637aab | 00:59 |
| clayg | now I just need a test | 00:59 |
| clayg | wait what the crap - why didn't test_hash_suffix_one_datafile already cover this? | 01:00 |
| clayg | oh what the hell verify defaults to False? | 01:03 |
| clayg | so how the.... | 01:03 |
| clayg | ok, I think maybe acoles already fixed the get_hashes thing | 01:09 |
| *** vinsh has joined #openstack-swift | 01:14 | |
| *** ozialien has quit IRC | 01:19 | |
| mattoliverau | Sorry, had a meeting | 01:23 |
| notmyname | clayg: so from what I can tell based on IRC buffers and gerrit reviews, acoles_away doesn't like the "frag-index" part of X-Backend-Ssync-Frag-Index and would prefer it to be archive-index | 01:27 |
| notmyname | but, the value (both in the object server and in ssync) is passed to "frag_index" in the diskfile. | 01:27 |
| clayg | maybe? maybe he doesn't like the ssync - assuming whatever we come up with might eventually get riffed on to handle the proxy server retrying requests with the specific goal of finding some missing fragment indexes? | 01:28 |
| notmyname | seems the issue comes from x-object-sysmeta-ARCHIVE-index vs x-backend-ssync-FRAG-index | 01:28 |
| clayg | hrmmm... i'm not reading his comment right now | 01:28 |
| notmyname | and he did say in here earlier that he's ok with how it is | 01:28 |
| clayg | so maybe we leave it? | 01:28 |
| clayg | i can't get this fucking unit test to fail the same way the probe test is failing | 01:29 |
| *** fanyaohong has quit IRC | 01:29 | |
| notmyname | well, if it changes, perhaps the change should be fore x-object-sysmeta-frag-index. ie keep it FRAG-index everwhere | 01:30 |
| notmyname | ugh. of course that's the one that has "42 matches across 9 files" | 01:30 |
| clayg | oh yeah - i rather like x-object-sysmeta-frag-index | 01:31 |
| clayg | and x-backend-frag-index (or maybe x-backend-ssync-frag-index) | 01:31 |
| notmyname | well I'd definitely agree that it's not good to have frag one place and archive somewhere else. no need to have multiple words for the same thing when one will do | 01:32 |
| clayg | i like x-backend-if-not-match-frag-indexes: <csv-of-frag-indexes-proxy-already-has> | 01:32 |
| clayg | sweet! | 01:32 |
| clayg | frag frag frag | 01:32 |
| notmyname | ok, I'll try a global find/replace to make it frag-index | 01:32 |
| clayg | it's like I'm playing quake1 and I've got the rl and the ra on lockdown in dm6 | 01:32 |
| notmyname | clayg: is x-backend-if-not-match-frag-indexes actually a thing? | 01:32 |
| clayg | notmyname: only in the mind's eye of torgomatic | 01:33 |
| notmyname | EC KILLING SPREE | 01:33 |
| * clayg totally wants to play q1 now | 01:33 | |
| mattoliverau | what could possibly go wrong | 01:33 |
| *** PurpleJack has joined #openstack-swift | 01:33 | |
| clayg | oh man team dm on dm3 - we used to totally control the quad - pwn fools | 01:33 |
| notmyname | clayg: so I saw the trailer for the new mortal combat game. reminded me of the first one, which I think I played at your house, and how much my parents hated it. of course, I looked at the new one and though "that's terrible. no way my kids are ever playing that!" | 01:34 |
| clayg | lol! *perspective* | 01:34 |
| swifterdarrell | notmyname: lol | 01:34 |
| notmyname | no kidding! | 01:34 |
| clayg | peluse: why is there a POST in here?! | 01:37 |
| *** PurpleJack has quit IRC | 01:38 | |
| clayg | does fast-post even work with ssync!? when did acoles_away fix THAT? | 01:38 |
| *** jrichli has quit IRC | 01:38 | |
| clayg | peluse: all the reconstructor tests have posts in them - did totally looks like you did it on purpose - are you running object_post_as_copy with true or false? | 01:39 |
| notmyname | clayg: did I see a hint that the current ec_Review branch doesn't pass probe tests? | 01:42 |
| clayg | ummm.... no i'm not sure that's true - peluse has a new test that was failing - when I ran them on friday it was working - almost positive | 01:43 |
| notmyname | ah ok | 01:43 |
| notmyname | must have been peluse saying something then | 01:43 |
| notmyname | ok. running tests locally, then will test the patch, then will put the patch up for you somewhere | 01:44 |
| clayg | owww, my mind is full of fuuuu | 01:46 |
| *** jrichli has joined #openstack-swift | 01:46 | |
| clayg | MIND == BLOWN | 01:51 |
| clayg | i don't know if i can do this without acoles :'( | 01:52 |
| *** erlon has quit IRC | 02:01 | |
| peluse | OK, I'm back | 02:05 |
| clayg | peluse: justify these POSTs all up in the probetests! | 02:05 |
| mattoliverau | clayg: so where you at? Maybe if you push where your up to we can take a look and combine brain power | 02:06 |
| peluse | was messing around with metadata and never removed them | 02:06 |
| clayg | mattoliverau: well... where I'm at is sorta broken - thanks to peluse | 02:06 |
| peluse | :) | 02:06 |
| peluse | so what's busted? | 02:07 |
| clayg | I also feel like torgomatic was going to send a diff - that I think I never saw | 02:07 |
| clayg | notmyname: may also have something I need to look at | 02:07 |
| clayg | peluse: a PUT to an object with a .data and .meta (no durable) - raises assertion error instead of diskfilenotexist | 02:07 |
| peluse | hmmm, how'd I have anything to do with that and when did it stop working? | 02:08 |
| clayg | peluse: near as I can tell it's always been this way - it's normally find - then you wrote a probe test to validate missing durables and all hell broke loose | 02:08 |
| clayg | peluse: why to write good tests! | 02:08 |
| peluse | heh, OK. sometimes even a blind squirrel finds a nut | 02:09 |
| peluse | so where do you need me at the moment? | 02:09 |
| mattoliverau | clayg: how about you give a git diff gist of where your currently at so we can all be on the same page, which patch are you on, reconstructor? | 02:09 |
| clayg | mattoliverau: I think the bug I'm working on would fold into the diskfile patch acctually - but it's a reconstructor bug | 02:11 |
| peluse | clayg, OK, so I'm going to beat on the reconstructor patch some more til I hear otherwise.... | 02:13 |
| clayg | which one? | 02:13 |
| clayg | mattoliverau: you're right I have too much code checked out for anyone else to make progress - let me try to get it up - everone take 5 | 02:14 |
| peluse | just the latest | 02:14 |
| peluse | ahh, OK | 02:14 |
| clayg | ok, i'm not even sure all those tests pass - much less pep8 worries | 02:21 |
| mattoliverau | clayg: that's fine for now :) | 02:21 |
| *** bkopilov has quit IRC | 02:22 | |
| clayg | anyway i'm zeroing in the the unittest failure that describes the issue i'm seeing with get_ondisk_files or _get_ondisk_file or gather_ondisk_files or how ever you want to get to it | 02:22 |
| notmyname | how did I get permission denied on probetests?! /me apparently doesn't know how to saio | 02:22 |
| mattoliverau | notmyname: saio properly damn it :P | 02:23 |
| clayg | peluse: oh, i wanted you to confirm if you're running with post_as_copy = false on your setup? since you added the client post to all the reconstructor test | 02:23 |
| notmyname | mattoliverau: I'm trying! I'm trying | 02:23 |
| clayg | notmyname: maybe it has something to do with the way the reconstrctuor tests simulate unmounted disks? | 02:23 |
| peluse | checking.... | 02:23 |
| clayg | notmyname: which probe tests fail? | 02:23 |
| notmyname | test_revert_object in /home/swift/swift/test/probe/test_reconstructor_revert.py | 02:24 |
| notmyname | renamer | 02:24 |
| clayg | peluse: if you're getting .meta files you've got post_as_copy = false - and you would be seeing the same failure I am after you fixed the suffix delta thing (or pull down my latest push) | 02:24 |
| peluse | notmyname, you should look at some of the Swift 101 materials from SwiftStack, thy're pretty good | 02:24 |
| clayg | which *does* surprising past unittests at the end of the chain btw - not sure that's saying much | 02:24 |
| peluse | what's the default for post_as_copy, that's what I have | 02:24 |
| notmyname | peluse: do they have EC docs yet? 'cause that would make this easier ;-) | 02:24 |
| notmyname | peluse: default is true | 02:24 |
| peluse | clayg, , OK mine is set to true :) | 02:25 |
| *** jrichli has quit IRC | 02:25 | |
| peluse | notmyname, yeah to simulate unmounted disks in the revert test, dirs are renamed to make them appear gone | 02:27 |
| notmyname | ah that woudl do it | 02:27 |
| clayg | mattoliverau: peluse: notmyname: so this is closest I can get to a test that seems to demonstrate what I think should be a problem -> https://gist.github.com/clayg/d47fba5611e9eeeec5b2 | 02:28 |
| notmyname | parent dir is owned by root | 02:28 |
| peluse | notmyname, because there are multiple devices behind each object server using the probetest kill option to simulate was taking out 2 at a time | 02:28 |
| clayg | basically down at the bottom I i have two assertRaises - the first one works - a datafile with no .durable acts exactly as if the .data isn't there | 02:28 |
| clayg | but the second one - the one that doesn't send in frag_index - like the object server would do - raises an assertion error | 02:28 |
| clayg | so somewhere in acoles visitor pattern code is a mean sneaky difference in behavior depending on if you passed in the frag_index - and I need it to just work even when the object server doesn't pass in the frag index - or else a missing durable is hella annoying | 02:29 |
| clayg | torgomatic: ARE YOu STILL here? | 02:29 |
| clayg | oh, nm it was just acorwin | 02:30 |
| *** thumpba has quit IRC | 02:30 | |
| clayg | i'm going to let you guys stew on that and see if peluse's probetest that started all this pain will pass with post_as_copy=false | 02:30 |
| peluse | OK, stewing | 02:31 |
| clayg | pthththth - well that was easy :D | 02:32 |
| clayg | ... still I'm pretty sure the diskfile issue is more low level than all that - the problem was ssync and .meta (known issues) - it's really that a PUT will try to read existing metadata - and since it doesn't hand in a frag_index - and we get different behavior depending on if you hand in a frag index or not - and the behavior we want is when you hand in frag_index - and we don't hand in frag_index - it sucks | 02:33 |
| *** ho has quit IRC | 02:34 | |
| *** ho has joined #openstack-swift | 02:34 | |
| clayg | you know what it is | 02:35 |
| peluse | not yet, do you? | 02:35 |
| clayg | _gather_on_disk_file has a dangling elif - gotta be the issue | 02:35 |
| clayg | ;) | 02:35 |
| clayg | i guess i need to trace it - somehow the if frag_index is None in the .data block of the if durable is tripping it up | 02:36 |
| peluse | yeah, I'm walking through it now | 02:37 |
| mattoliverau | lol | 02:37 |
| clayg | a smaller unittest would be nice - there's one called test_get_ondisk_files_with_stray_meta that is proving it raises a assertionerror - and I think it's basically wrong or something - or maybe if you gave it frag_index=1 it would fail because you'd get DiskFileDoesNotExist instead (which is what we want btw) | 02:37 |
| clayg | notmyname: honestly I'm not even sure this is as important as the quorum cleanup that torgomatic was working on - but I'm not sure where the hammer fell down on that | 02:38 |
| clayg | I think basically torgomatic decided the existing code worked, but could be more obvious that it works - because acoles_away and peluse both seemed to agree that despite the tests proving it works - it looks to broken to possibly be correct | 02:39 |
| torgomatic | clayg: notmyname: basically there's one branch that's obviously wrong, but that's okay (LOL) because it's dead code | 02:39 |
| clayg | I gave up on understandinng multi-phase put ages ago and have been relying on the tests almost entirely - but I didn't try to write a test specifically to demonstrate the issue they're concerned about | 02:39 |
| torgomatic | or at least, not called | 02:39 |
| torgomatic | you have to call _get_put_responses with thing1=True and thing2=True, and the only two call sites are F/T and T/F | 02:40 |
| clayg | torgomatic: well can you give me a diff that removes it? | 02:40 |
| clayg | torgomatic: I mean... if you have one | 02:40 |
| torgomatic | clayg: https://gist.github.com/smerritt/ceacbc30d29772b52e51 | 02:40 |
| clayg | yay! | 02:40 |
| torgomatic | tests are still running on my machine though, so give me about 300 seconds | 02:41 |
| clayg | torgomatic: ack! thanks! | 02:41 |
| notmyname | (same here for fragfragfrag) | 02:41 |
| clayg | well then everyone stop worrying! geez | 02:42 |
| torgomatic | fragfragfrag? | 02:42 |
| notmyname | you'll see | 02:42 |
| * torgomatic hides | 02:42 | |
| notmyname | also quake1 | 02:42 |
| clayg | brb | 02:42 |
| notmyname | torgomatic: conversation between clayg and acoles_away earlier based on something acoles_away said in gerrit. changing the name of one of the headers to be consistent | 02:43 |
| torgomatic | ok, my tests pass | 02:45 |
| *** annegentle has joined #openstack-swift | 02:46 | |
| *** zackmdavis has quit IRC | 02:48 | |
| *** zackmdavis has joined #openstack-swift | 02:49 | |
| *** annegentle has quit IRC | 02:55 | |
| clayg | torgomatic: ok so i should squash that diff you gave - or did you add anything else? | 02:58 |
| torgomatic | clayg: that's it, really | 02:58 |
| clayg | mattoliverau: peluse: any luck with the gather on diskf iles? | 02:58 |
| torgomatic | there's one other tiny cleanup, but it doesn't affect correctness | 02:58 |
| mattoliverau | clayg: df.write_metadata(extra_meta) fails in your new test with an assert error 'On-disk file search algorithm contract is broken' because it doesn't find the durable file. So can't get a successful even first assert | 02:58 |
| peluse | clayg, at the top of _verify_on_disk_files() there's a check for "sole meta" that checks for fi not equal non, if you remove that check ithe test passes | 02:58 |
| peluse | if frag_index is not None and not accepted_files.get('.data'): | 02:58 |
| peluse | # We may find only a .meta if searching for a specific frag index | 02:58 |
| peluse | # that is not present. That does not mean the on disk contract is | 02:58 |
| peluse | # broken, so clear the .meta to comply with superclass assertions. | 02:58 |
| peluse | accepted_files['.meta'] = None | 02:58 |
| peluse | so the comment suggests acoles_away felt like this was only valid if FI was specified | 02:59 |
| clayg | peluse: where was that? | 03:00 |
| peluse | _verify_on_disk_files() | 03:00 |
| clayg | ahh _verify | 03:00 |
| peluse | if you kill the "if frag_index is not None" condition things work | 03:00 |
| clayg | peluse: do any other tests fail? | 03:00 |
| peluse | good Q, one min | 03:00 |
| clayg | because I really like the idea of just clearing the .meta if not .data | 03:01 |
| peluse | yeh, a few :) | 03:01 |
| *** wolsen_ has joined #openstack-swift | 03:01 | |
| peluse | 2 errors, 3 failures | 03:01 |
| clayg | hrmmm... i only saw the two failures | 03:02 |
| peluse | just assertions to be tweaked because of the code change though | 03:02 |
| clayg | peluse: this was my interpretation of your words -> https://gist.github.com/clayg/07468e4c5476b87e50e2 | 03:03 |
| clayg | and both of the failures I see look totally reasonable in light of the behavior we want | 03:03 |
| peluse | yup | 03:03 |
| peluse | agree | 03:03 |
| peluse | and I believe this is just an overlooked case and that change is a good fix | 03:03 |
| peluse | w/the comment cleaned up of course | 03:04 |
| mattoliverau | nice work guys | 03:04 |
| clayg | oh yeah the comment | 03:04 |
| peluse | because clearly we've found a case where you can have .meta without the .durable :) | 03:04 |
| clayg | e.g. one of peluse's terribly terribly mean probetests | 03:05 |
| notmyname | clayg: https://gist.github.com/notmyname/effac114a81e5d1bbeca | 03:06 |
| *** zackmdavis has quit IRC | 03:06 | |
| *** zaitcev has quit IRC | 03:06 | |
| *** Trixboxer has quit IRC | 03:06 | |
| *** nottrobin has quit IRC | 03:06 | |
| *** goodes has quit IRC | 03:06 | |
| *** wolsen has quit IRC | 03:06 | |
| *** tristanC has quit IRC | 03:06 | |
| notmyname | clayg: that works, but I need to figure out probetests. after chowning the dirs, i'm getting new and exciting errors | 03:06 |
| *** silor has joined #openstack-swift | 03:07 | |
| notmyname | clayg: that's a diff on top of the docs patch (end of the chain) | 03:07 |
| mattoliverau | notmyname: you could probably change the comment "+* ``X-Object-Sysmeta-Ec-Frag-Index``: Also known as the "fragment index"." cause obviously frag-index is known as the fragment index, I think it said that before to let people know that archive-index is also known as fragment index. | 03:08 |
| mattoliverau | but meh, clayg can fix that :P | 03:08 |
| clayg | mattoliverau: thanks :\ | 03:09 |
| notmyname | lol | 03:09 |
| clayg | peluse: gd, I have no idea how to fix either of these tests | 03:09 |
| mattoliverau | clayg: you are the ec_review master after all :P | 03:09 |
| clayg | the 'make sure this breaks' assertion error test is crap - it's basically - make sure there's a landmind here - and we're like - we need to fix this landmine! | 03:09 |
| peluse | heh, brb | 03:11 |
| notmyname | https://gist.github.com/notmyname/9fd0a28fa36b9fe9c1de | 03:12 |
| notmyname | clayg: ^^ better version with better words for docs | 03:12 |
| notmyname | mattoliverau: there | 03:13 |
| *** tristanC has joined #openstack-swift | 03:13 | |
| *** zackmdavis has joined #openstack-swift | 03:13 | |
| *** zaitcev has joined #openstack-swift | 03:13 | |
| *** Trixboxer has joined #openstack-swift | 03:13 | |
| *** nottrobin has joined #openstack-swift | 03:13 | |
| *** goodes has joined #openstack-swift | 03:13 | |
| *** sendak.freenode.net sets mode: +v zaitcev | 03:13 | |
| notmyname | torgomatic: that's fragfragfrag | 03:13 |
| mattoliverau | notmyname: much better ;) | 03:13 |
| clayg | notmyname: next, can you address acoles last comment on -> https://review.openstack.org/#/c/169990/ | 03:14 |
| mattoliverau | I call that reviewing in flight (before it hit gerrit) :P | 03:14 |
| notmyname | clayg: this seems apropos at the moment https://twitter.com/teabass/status/585436639423569921 | 03:14 |
| clayg | peluse: I'm inclined to this fix for the assertion test -> https://gist.github.com/clayg/294e0399f335296e57a6 | 03:14 |
| clayg | peluse: still not sure what to do with the gather ondisk files failure for the scenario I added of .data .meta (no durable) | 03:15 |
| notmyname | clayg: yes. on it. and I need to take his comments for the changelog too along the same lines | 03:15 |
| * peluse on the phone for a few minutes | 03:15 | |
| clayg | notmyname: good one | 03:16 |
| clayg | ok, found the solution for the gather_tests too - running the mean probe test | 03:18 |
| clayg | notmyname: oh... maybe you should figure out what's going on with your probetests instead | 03:18 |
| clayg | mattoliverau: you're a word smith - can you do anything about acoles comment on the doc patch? | 03:18 |
| notmyname | probably because I use real disks | 03:18 |
| mattoliverau | clayg: I'll take a look | 03:19 |
| notmyname | clayg: mattoliverau: I've got the docs change | 03:19 |
| clayg | notmyname: so what are the things at /srv/nodeX/sdbX - symlinks? | 03:19 |
| mattoliverau | cool | 03:19 |
| clayg | notmyname: I want to say all peluse does is rename them - are they not owned by the user running tests maybe? | 03:19 |
| notmyname | clayg: https://gist.github.com/notmyname/04f4e8c04c7c24ec9e13 | 03:19 |
| clayg | oh | 03:20 |
| clayg | ok, so - i acctually have a diff that would help with that sorta | 03:20 |
| clayg | well not entirely | 03:20 |
| clayg | it makes helper methods to do the "make this disk act like it's unmounted" | 03:21 |
| notmyname | oh cool | 03:21 |
| clayg | so if we used those helper methods in all the reconstrctor tests - and then made them smart in the face of os.ismount - that would be slightly better | 03:21 |
| notmyname | I'm not in love with the saio config of having mount points like that. doubling up on 1..4 | 03:21 |
| clayg | in that you know... our ptl could acctually run probetests :\ | 03:21 |
| *** PurpleJack has joined #openstack-swift | 03:22 | |
| clayg | notmyname: not sure I follow? having more than one disk per object-server process is bumming you out? or just the way that it goes 1&4,2&5,3&6.. | 03:22 |
| notmyname | oh, yeah that's fine. /undo previous comment | 03:23 |
| clayg | bah, nm the diff i had for the brain splitter decided to "fix" the account & container servers (via BaseStorageServer no less) to work with the check_dir constraint like the object server does | 03:24 |
| clayg | no wonder I decided to throw it in a branch until after EC | 03:24 |
| clayg | THIS IS WHY WE CAN'T HAVE NICE THINGS | 03:25 |
| clayg | notmyname: so it's just the revert test that is failing for you? | 03:26 |
| notmyname | yup | 03:27 |
| *** PurpleJack has quit IRC | 03:27 | |
| notmyname | mattoliverau: wordsmithing https://etherpad.openstack.org/p/wJULuIYgXc | 03:27 |
| clayg | wow, i don't think i get the same page you do there :\ | 03:29 |
| notmyname | I'm told a refresh seems to make it sort itself out | 03:29 |
| notmyname | -infra just upgraded the etherpad install | 03:29 |
| clayg | notmyname: how are we going to programmatically unmount and mount your devices? does os.system('umount %s' % device_path) work? | 03:29 |
| clayg | maybe too much.. | 03:30 |
| notmyname | "only root can unmount..." | 03:30 |
| clayg | notmyname: well gd - what's a test supposed to do!? | 03:31 |
| peluse | oh man, this is for notmynames non-SAIO env? | 03:31 |
| clayg | heh... what's a test to *SU*do???? | 03:31 |
| notmyname | peluse: no, my saio env! | 03:31 |
| peluse | yeah, I mean using real moint points :) | 03:31 |
| clayg | I think torgomatic may also roll with mount_check=True - maybe he uses symlinks tho | 03:31 |
| clayg | notmyname: does the docs say to use symlinks? | 03:32 |
| torgomatic | my stuff's all symlinky | 03:32 |
| notmyname | torgomatic: teach me! | 03:32 |
| torgomatic | the docs do recommend symlinks, I think | 03:32 |
| * notmyname goes and looks | 03:32 | |
| clayg | peluse: good news - durable probetest thing passes with our diskfile changes - you rock! | 03:32 |
| notmyname | clayg: I'll be it works AWESOME with vagrant-saio | 03:32 |
| peluse | sweet | 03:33 |
| clayg | notmyname: of course it does | 03:33 |
| torgomatic | notmyname: I dunno; I just copied and pasted what was in the docs | 03:33 |
| torgomatic | make a loopback xfs thingy, then make a bunch of symlinks into it | 03:33 |
| clayg | torgomatic: oh so you *do* use mount_check=false | 03:33 |
| clayg | notmyname: I think I honestly used to do what you have - direct mounts of real devices into /srv/node | 03:34 |
| torgomatic | clayg: yeah, looks like it | 03:34 |
| clayg | which I think it'd be nice if the probetests supported | 03:34 |
| clayg | notmyname: can you try os.system('sudo unmount %s' % device_path) ??? | 03:34 |
| notmyname | clayg: totally works (since my swift user is ALL=NOPASSWD: ALL) | 03:35 |
| clayg | notmyname: oh man - your frag changes is going to be all over the diffs | 03:35 |
| clayg | ok, I think i can work with that | 03:36 |
| notmyname | mattoliverau: note updated. look good? | 03:37 |
| peluse | notmyname, yeah, I added that note. Looks good | 03:38 |
| notmyname | peluse: ah, thanks :-) | 03:38 |
| peluse | notmyname, do you want to specifically mention performance though just in case its a real dog? | 03:38 |
| notmyname | clayg: are you about to do something? can I push over the docs patch? | 03:39 |
| mattoliverau | notmyname: yup, looks good to me, and should now pop in the documentation :) | 03:39 |
| notmyname | peluse: no, not in that note. perf is mentioned in other places | 03:39 |
| peluse | cool | 03:39 |
| peluse | wanted to make sure it wasn't overlooked | 03:39 |
| clayg | notmyname: you can push | 03:40 |
| notmyname | peluse: I'll add a note into the "performance considerations" section later in that doc | 03:40 |
| peluse | col | 03:40 |
| peluse | l | 03:40 |
| peluse | cool | 03:40 |
| peluse | jesus | 03:40 |
| clayg | heh | 03:40 |
| mattoliverau | lol | 03:40 |
| clayg | peluse: do you want to try to extra the calls to renamer in test_revert to call some helpers called "unmount" and "mount" so that notmyname and I can play around with an if os.ismount check branch that will call os.system() instead of the renamer? | 03:41 |
| clayg | s/extra/extract/ | 03:41 |
| peluse | sure | 03:41 |
| clayg | peluse: i'm seeing if I can get all these changes from notmyname's diff screwed into the right patchsets | 03:42 |
| clayg | peluse: you're a doll! | 03:42 |
| notmyname | peluse: | 03:42 |
| * clayg needs some tunes | 03:42 | |
| notmyname | In general, EC has different performance characteristics than replicated data. | 03:42 |
| notmyname | EC requires substantially more CPU to read and write data, and is more suited | 03:42 |
| notmyname | to larger objects that are not frequently accessed (eg backups). | 03:42 |
| peluse | oh ow, that's a first! | 03:42 |
| notmyname | is that paragraph sufficient for now for perf considerations? | 03:43 |
| notmyname | s/suited to larger/suited for larger/ | 03:43 |
| clayg | notmyname: read/write/reconstruct/reocvery/consistency/something | 03:44 |
| notmyname | clayg: *everything*. you know what? just go by a pallet of CPUs before you deploy this | 03:44 |
| clayg | ... probably :\ | 03:44 |
| mattoliverau | Visit Intel at: URL :p | 03:45 |
| notmyname | lol | 03:45 |
| clayg | also infinite bandwith and all flash primary storage | 03:45 |
| clayg | ^ would be nice | 03:45 |
| notmyname | clayg: ok, docs just pushed | 03:46 |
| clayg | notmyname: cool | 03:46 |
| * torgomatic goes to put the kid to bed | 03:49 | |
| openstackgerrit | John Dickinson proposed openstack/swift: 2.3.0 authors and changelog updates https://review.openstack.org/172573 | 03:52 |
| notmyname | same words put there | 03:52 |
| clayg | peluse: oh wait - did you have a fixup for the _verify method that included the better words for the comment? | 03:52 |
| peluse | no, just remove the part about "if looking for specific FI" | 03:53 |
| peluse | is this what you were thinking for the revert change? https://gist.github.com/peluse/54b2b266b2856075e2f6 | 03:54 |
| clayg | peluse: you haven't been paying attention | 03:54 |
| peluse | and then add whatever switch for doing mount/unmount insttead | 03:54 |
| clayg | any time I try to type a comment or doc string it's a terrible mess of spelling errors and fixups | 03:54 |
| clayg | the only hope is a diff | 03:54 |
| peluse | yeah, like my typing is any beter | 03:54 |
| sorrison | hi guys have another swift issue for you I'm hoping you can help me with | 03:54 |
| clayg | peluse: well maybe if you can get mattoliverau to review it | 03:54 |
| notmyname | sorrison: pm me about it, please | 03:55 |
| sorrison | sweet | 03:55 |
| peluse | yeah, mattoliverau can type! | 03:55 |
| clayg | peluse: yeah mostly like that - except that instead of if self.kill_method it'll be a if os.ismount | 03:55 |
| clayg | peluse: anyway - i'll take that diff - thanks | 03:55 |
| *** a7ndrew has joined #openstack-swift | 03:56 | |
| peluse | well, I figured you'd put that in the kill_drive() routine based on kill_method as defined by someone for some reason | 03:56 |
| mattoliverau | what am I typing? rewording the verify comment? | 03:56 |
| peluse | yeah, that's the ticket | 03:56 |
| clayg | mattoliverau: oh yess please! | 03:56 |
| peluse | oh, os.ismount, duh. that's the condition | 03:57 |
| clayg | oh right, I was still auditing acoles comments on the reconstructor change | 03:58 |
| clayg | peluse: you're not doing anything right now right? | 03:58 |
| clayg | peluse: basically i just go through the comments on patch set 6 and make sure they all get marked Done after verifying the fix in patch set 7 or make new comments on patch set 7 | 03:59 |
| peluse | sure | 04:00 |
| peluse | you mean you want me to do that right? | 04:00 |
| clayg | unless you had something else in mind :D | 04:02 |
| *** bkopilov has joined #openstack-swift | 04:02 | |
| peluse | nope, I'm on it | 04:03 |
| peluse | did you mean to include the new probe test for missing durable in that set or is another pending? | 04:03 |
| clayg | another is pending - now that it passes I'm going to add it | 04:04 |
| clayg | mattoliverau: you got some words for me? | 04:05 |
| peluse | cool, I'll get this one all audited in the meantime | 04:05 |
| mattoliverau | clayg: something like: http://paste.openstack.org/show/203817/ | 04:05 |
| clayg | poetry | 04:05 |
| mattoliverau | right, just gabbing some lunch brb | 04:07 |
| clayg | notmyname: I thought the Co-Authoried-By lines were supposed to be alpha betical - but that's not true | 04:08 |
| notmyname | oh? | 04:08 |
| notmyname | I thought so | 04:09 |
| notmyname | did I mess it up? | 04:09 |
| notmyname | doh! I always mess up ordering with tdasilva | 04:09 |
| clayg | notmyname: oh nm, I apparently don't know what letter da Silva starts with | 04:09 |
| notmyname | my eyes go to the "S" and he uses the "da" | 04:09 |
| notmyname | it's still wrong :-) | 04:10 |
| clayg | are you sure? does it "startwith" d or S? | 04:10 |
| notmyname | da is before di | 04:12 |
| clayg | and it has come to this -> http://en.wikipedia.org/wiki/Alphabetical_order#Ordering_by_surname | 04:12 |
| notmyname | that says, "good luck" | 04:13 |
| notmyname | clayg: just put his name before mine. that's the right way. | 04:14 |
| *** a7ndrew has left #openstack-swift | 04:14 | |
| clayg | notmyname: John Dickinson <me@not.mn> ? | 04:14 |
| notmyname | yup. oh yeah. looks like I added a : there too. don't do that | 04:15 |
| notmyname | again, http://i.imgur.com/ES6sP.jpg | 04:16 |
| peluse | clayg, OK, done | 04:16 |
| clayg | hehehe - not quite as funny the second time - but still awesome | 04:17 |
| clayg | peluse: thanks! | 04:17 |
| peluse | my pleasure!! | 04:17 |
| clayg | peluse: did I miss anything important or mostly just acoles_away whining? | 04:17 |
| clayg | peluse: oh right the etag thing - i acctually had no idea what he was talking about there :\ | 04:18 |
| peluse | nothing important except maybe the one about Etag vs ETag in the odl DiskFileLikeThingy class that you renamed to something really professional soundsing :) | 04:18 |
| peluse | yeah, me neither! | 04:18 |
| clayg | lol at something professional | 04:19 |
| clayg | fuck that noise | 04:19 |
| clayg | ;) | 04:19 |
| *** thumpba has joined #openstack-swift | 04:19 | |
| peluse | oh, and this. I'll add a comment: test/probe/test_reconstructor_prop_durable.py:20:1: F401 'shutil' imported but unused | 04:19 |
| openstackgerrit | John Dickinson proposed openstack/swift: 2.3.0 authors and changelog updates https://review.openstack.org/172573 | 04:19 |
| clayg | doh! | 04:19 |
| peluse | oh wait, you haven't taken that file in yet :) | 04:19 |
| clayg | I think i just named it test_reconstructor_durable - but I did remove the shutil ;) | 04:19 |
| clayg | bah - i guess i'll try to understand what he was getting at with the etag thing :\ | 04:20 |
| peluse | but when you do.... https://gist.github.com/peluse/a38f8fb516425ef2a905 | 04:20 |
| peluse | oh, OK, never mind :) | 04:21 |
| clayg | peluse: i think that test does not use proxy_get? | 04:21 |
| peluse | yeah, I dont think so | 04:21 |
| peluse | nope | 04:22 |
| peluse | clayg, I'm done for the night... leave me a note for a.m. work I'll be on at 5ish ready to do whatever | 04:24 |
| notmyname | peluse: should be done by then, I hope | 04:24 |
| clayg | peluse: ok, np thanks for everything | 04:25 |
| peluse | or if there's something I can do now - say the word | 04:25 |
| clayg | yeah we're bout to pull this thing out of the oven | 04:25 |
| clayg | god i wish flake8 wasn't so fucking slow - why is flake8 so fucking slow? | 04:25 |
| notmyname | clayg: you've been rebasing ec_review against master right? | 04:31 |
| notmyname | I think there was one thing that landed there | 04:31 |
| clayg | notmyname: ummm... nope | 04:31 |
| notmyname | hmm | 04:31 |
| notmyname | looks like there might be a few conflicts. I'm going through them | 04:32 |
| clayg | awww man - can you get them pushed into ec_review? | 04:32 |
| notmyname | yes | 04:32 |
| clayg | gd, I think flake8 is slower than our unitests? how is this even possible!? | 04:32 |
| clayg | i think it's all those hacking checks | 04:32 |
| notmyname | wow. don't we have most of them turned off? | 04:33 |
| clayg | i bet it still runs them | 04:33 |
| clayg | then ignores the | 04:33 |
| clayg | m | 04:33 |
| notmyname | jsut supresses the result? | 04:33 |
| clayg | i don't really know | 04:33 |
| mattoliverau | And back | 04:33 |
| clayg | i'm just in a mood | 04:33 |
| clayg | mattoliverau: can you make flake8 faster? kthnxbai | 04:33 |
| mattoliverau | Sure let's just make it yay less :p | 04:34 |
| *** krtaylor has quit IRC | 04:35 | |
| *** krtaylor has joined #openstack-swift | 04:39 | |
| notmyname | clayg: what do you have queued to push? | 04:39 |
| notmyname | if it's several things, push it and I'll start the rebase from there | 04:39 |
| clayg | notmyname: but i'm on my last sha for my comprehensive flake8 test1 | 04:40 |
| notmyname | ok :-) | 04:40 |
| notmyname | hurry up, flake8!!!! | 04:40 |
| clayg | srly | 04:40 |
| *** ppai has joined #openstack-swift | 04:41 | |
| clayg | some how it seems unresonable to me that running flake8 on our code base ten times is a 20 minute affiar :\ | 04:41 |
| * notmyname goes to look for some food* while flake8 finishes | 04:43 | |
| torgomatic | the asterisk has me curious | 04:44 |
| mattoliverau | Yeah, is it liquid food? ;) | 04:45 |
| notmyname | ice cream! | 04:46 |
| notmyname | mattoliverau: there will be plenty of time for that after all this is through ;-) | 04:47 |
| torgomatic | so, eventually liquid. got it. | 04:47 |
| notmyname | torgomatic: most food is ;-) | 04:47 |
| mattoliverau | Lol | 04:47 |
| clayg | notmyname: just a couple more spot checks - I think I'm done | 04:50 |
| notmyname | kk | 04:50 |
| clayg | ok, I'm not sure I can think of anything I forgot | 04:52 |
| clayg | i'm going to push up where we're at - there some good stuff i've only got on my local machine - so better to throw it out here i hope | 04:52 |
| notmyname | ok | 04:52 |
| clayg | oh... but wait - you said there's some rebase against master I'm going to ahve to deal with? | 04:53 |
| notmyname | clayg: ya, I was going to take the patch chain and do that once you push it up | 04:53 |
| notmyname | ...or you can ;-) | 04:53 |
| clayg | notmyname: also you should try out that revert probe test see if I was anywhere close to right | 04:54 |
| notmyname | oh, I'll run tests | 04:54 |
| notmyname | clayg: if you push that up, I'll work on the rebase, and if you could gather a "where we're at" status, that'd be swell | 04:54 |
| clayg | where we're at == done as far as I know | 04:55 |
| notmyname | whoa!! | 04:55 |
| notmyname | well, still need to rebase against master and check it out. small things, only AFAICT | 04:55 |
| notmyname | torgomatic: still around? | 04:56 |
| torgomatic | yup | 04:56 |
| clayg | torgomatic: it's a trap! | 04:57 |
| notmyname | ok, so if we have the "done" patch chain up for ec_review, then we've got torgomatic clayg mattoliverau and myself for marking +2,+A on everything. of course with other existing votes from acoles_away cschwede peluse and tdasilva totally counting too | 04:57 |
| torgomatic | IOW enough votes to kick things off | 04:58 |
| notmyname | to satisfy the process gods | 04:58 |
| clayg | notmyname: i'm reviewing the diffs that came up from tonight to make sure i pushed up what I thought I did | 04:58 |
| notmyname | ack | 04:58 |
| clayg | i'm on the proxy change now - things are looking good | 04:59 |
| clayg | it's all s/archive/frag and fixing the gather-ondisk-assertion bug so far | 04:59 |
| notmyname | I was fully expecting to be up well past midnight. this is great news (/me counts chickens, "one egg, two eggs, ...") | 04:59 |
| clayg | oh and torgomatic's clean up of the quorum thing | 04:59 |
| notmyname | clayg: wait did you just push stuff? I got a bunch of emails | 05:00 |
| notmyname | did you rebase? | 05:01 |
| clayg | yeah sometime right after - "better to throw it out there" and "were at done" | 05:01 |
| clayg | notmyname: NO? | 05:01 |
| notmyname | ok. looking | 05:01 |
| clayg | was i suppose to? | 05:01 |
| torgomatic | possibly the frag stuff? | 05:01 |
| clayg | i'm confused | 05:01 |
| clayg | torgomatic: no i got all that notmyname sent me the diff | 05:01 |
| clayg | torgomatic: he's all up in your co-authors now! | 05:02 |
| torgomatic | clayg: ah, got it | 05:02 |
| notmyname | yeah, some merge conflicts with master. let me work through them | 05:02 |
| clayg | notmyname: sounds great! I'll get a beer! | 05:02 |
| clayg | s/a/another/g | 05:02 |
| notmyname | rebased and unit tests passed. running functests | 05:09 |
| notmyname | https://review.openstack.org/#/c/170340/ is the only one that conflicted (container sync one) | 05:10 |
| notmyname | diff is https://gist.github.com/notmyname/d3d86c8f46ea01585b9f | 05:10 |
| *** PurpleJack has joined #openstack-swift | 05:11 | |
| *** wer has quit IRC | 05:12 | |
| torgomatic | that looks like an easy enough merge | 05:13 |
| notmyname | clayg: did you intentionally leave off the co-authored-lines on https://review.openstack.org/#/c/170666/ and https://review.openstack.org/#/c/170340/ ? | 05:13 |
| *** ppai has quit IRC | 05:14 | |
| *** PurpleJack has quit IRC | 05:16 | |
| *** thumpba_ has joined #openstack-swift | 05:16 | |
| notmyname | clayg: hang on before you start editing, I think I can do it from here, if that's ok with you | 05:17 |
| clayg | notmyname: myabe? was that container sync and reaper? | 05:17 |
| notmyname | yes | 05:17 |
| clayg | notmyname: yeah knock yourself out! | 05:17 |
| *** thumpba has quit IRC | 05:17 | |
| *** wer has joined #openstack-swift | 05:17 | |
| clayg | i think container-sync was like all yuan - and the reaper I just threw together at some point - not point blaming everyone for that mess | 05:17 |
| notmyname | ok, ack | 05:17 |
| clayg | but I wouldn't be offended if that was "fixed" | 05:18 |
| *** SkyRocknRoll has joined #openstack-swift | 05:19 | |
| *** welldannit has quit IRC | 05:19 | |
| clayg | notmyname: anyway that diff for container sync looks right I think? | 05:19 |
| *** zaitcev has quit IRC | 05:20 | |
| notmyname | clayg: I'm assuming you added me only on the ones that I touched? | 05:20 |
| clayg | yeah I think so? | 05:20 |
| notmyname | ok, I assumed so. just checking | 05:20 |
| clayg | notmyname: hey test that revert prbetest for me | 05:21 |
| notmyname | running right now | 05:21 |
| *** thumpba_ has quit IRC | 05:23 | |
| clayg | mattoliverau: I wonder if the commit message for the ec-docs change should be better? https://review.openstack.org/#/c/169990 | 05:24 |
| clayg | notmyname: i just finished my spot check on the diffs I pushed tonight and I think they're all solid - just what we wanted to change - nothing slipped in that I saw | 05:25 |
| mattoliverau | yeah probably should be, I'll add one and push it up. | 05:25 |
| clayg | mattoliverau: co-ordinate the push with notmyname | 05:25 |
| notmyname | ok, great | 05:25 |
| notmyname | mattoliverau: ya, give me the text. I'm gonna push the whole chain up with some commit message stuff taken care of | 05:25 |
| mattoliverau | notmyname: sure, give me a minute then | 05:25 |
| notmyname | eg all the examples I saw had the co-authored lines in the last paragraph. | 05:26 |
| notmyname | also, I've got the rebased version | 05:26 |
| *** ppai has joined #openstack-swift | 05:28 | |
| notmyname | dooo-dee-doo /probetests/ | 05:28 |
| *** km has quit IRC | 05:29 | |
| *** km has joined #openstack-swift | 05:29 | |
| notmyname | oh, sorrison's issue he popped in to ask about earlier was that the remote region had been configured with different salts in swift.conf and therefore was quarantining everything | 05:30 |
| torgomatic | heh, I love that failure mode | 05:31 |
| torgomatic | "this one stupid machine keeps quarantining everything!" | 05:31 |
| notmyname | of course, the cluster seems fine to users :-) | 05:32 |
| mattoliverau | notmyname: http://paste.openstack.org/show/203820/ | 05:32 |
| cschwede | Good Morning! | 05:32 |
| notmyname | cschwede: yo! | 05:33 |
| mattoliverau | cschwede: morning! | 05:33 |
| notmyname | cschwede: just in time! | 05:33 |
| torgomatic | I wonder if we could also store the hash of a well-known string in obj meta somewhere, like hash_path("/openstack/swift") | 05:33 |
| kota_ | cschwede: morning | 05:33 |
| torgomatic | and then if that thing doesn't match, the auditor can emit a meaningful message | 05:33 |
| notmyname | oh wow. it is morning in germany | 05:33 |
| cschwede | it’s crowded in here! | 05:33 |
| torgomatic | that wouldn't take up too many bytes, especially if we gave it a short name and maybe even stored only like 4 bytes of hash | 05:34 |
| notmyname | cschwede: we're nearly done! | 05:34 |
| cschwede | notmyname: awesome :D | 05:35 |
| torgomatic | so you could still have undetected mistakes, but with low probability (on the order of 1 in 2**32) | 05:35 |
| clayg | torgomatic: you eamn like have the proxy send down the hash of that path with every object PUT? | 05:37 |
| torgomatic | clayg: no, just have the object server compute that and store it | 05:37 |
| torgomatic | although that might not help... yeah, I dunno | 05:37 |
| cschwede | mattoliverau: notmyname: theres an extra colon after Johns name in http://paste.openstack.org/show/203820/ | 05:38 |
| torgomatic | you're right, the proxy would have to do it | 05:38 |
| notmyname | cschwede: I'll fix it | 05:38 |
| clayg | torgomatic: I have no idea what we're talking about | 05:38 |
| mattoliverau | cschwede: damn those colons | 05:38 |
| clayg | torgomatic: why can't you just have a management system that always rights out a correct swift.conf ;) | 05:38 |
| torgomatic | clayg: some guy screwed up his swift.conf on one node, so his auditor quarantined the world | 05:38 |
| torgomatic | clayg: seriously ;) | 05:38 |
| notmyname | clayg: I'm still getting "OSError: [Errno 16] Device or resource busy" | 05:38 |
| *** proteusguy has quit IRC | 05:38 | |
| clayg | notmyname: well damnit! | 05:39 |
| cschwede | what is the most important thing this morning/evening, besides adding my +2s? | 05:39 |
| notmyname | clayg: trying different permutations: umount -f, umount -fl, etc | 05:39 |
| clayg | cschwede: notmyname doesn't know how to saio | 05:39 |
| clayg | lazy unmount is the worst :\ | 05:39 |
| clayg | i've found its almost never what I want - but YMMV | 05:39 |
| notmyname | what's the umount invocation to actually unmount DO IT NOW | 05:39 |
| notmyname | also, umount vs unmount? | 05:39 |
| clayg | notmyname: don't be snot - i told you I didn't test it | 05:40 |
| notmyname | no, I'd assumed I was wrong there ;-) | 05:40 |
| clayg | notmyname: heh, you guys only *thought* my code was crap *before* you realized this is only as bad as it gets *after* I wrote a bunch of unittests :\ | 05:41 |
| clayg | notmyname: my supply of bugs is practially unlimited when I don't have to test it | 05:42 |
| clayg | notmyname: how do you think I come up with our interview problems? | 05:42 |
| notmyname | so `umount -fl` seems to be the right spelling, but why doesn't it work? | 05:42 |
| * clayg types something up and does run it | 05:42 | |
| clayg | *BAM* crapy code | 05:42 |
| notmyname | "so we've got a guy who writes bad code and you need to fix it" <<-- actual swiftstack interview question | 05:43 |
| clayg | ^ say every person ever that's had to work with clayg | 05:43 |
| cschwede | interesting tactic! | 05:43 |
| clayg | notmyname: i'm not sure, lsof might have something interesting to say about it | 05:43 |
| notmyname | hang on | 05:44 |
| clayg | cschwede: so it's early for you ya? | 05:44 |
| clayg | cschwede: what's the delta with you and acoles_away ? | 05:45 |
| clayg | cschwede: all of europe has like one timezone right? | 05:45 |
| mattoliverau | which is different then the UK | 05:45 |
| cschwede | clayg: for me its 7:45am; I’m a little bit earlier than normal to catch up with you guys | 05:45 |
| notmyname | mattoliverau: no, that's the money | 05:45 |
| cschwede | acoles is one hour difference, 6:45am now | 05:45 |
| clayg | awwww | 05:45 |
| mattoliverau | lol | 05:45 |
| notmyname | oh! | 05:46 |
| clayg | mattoliverau: hey that's a good commit message | 05:46 |
| notmyname | I see what the issue is | 05:46 |
| notmyname | clayg didn't test it! ;-) | 05:47 |
| mattoliverau | clayg: well better then what was there :P | 05:47 |
| clayg | check-tempest-dsvm-full i'm going to punch you in the teeth | 05:47 |
| notmyname | got it | 05:47 |
| notmyname | now which patch is that in? | 05:48 |
| clayg | notmyname: ec-recon | 05:48 |
| notmyname | thanks | 05:48 |
| clayg | kota_: did you post comments to a stale diff? | 05:48 |
| kota_ | clayg: yes, I was reviewing during you changes something like frag-index | 05:49 |
| kota_ | clayg: for now, I 'm on the rebased one of the same patch (patch 169987) | 05:50 |
| patchbot | kota_: https://review.openstack.org/#/c/169987/ | 05:50 |
| notmyname | ok, I'm ready to push a "final" patch chain | 05:50 |
| notmyname | something from kota_ to look at? | 05:51 |
| clayg | kota_: http://effbot.org/zone/default-values.htm <- interesting read | 05:51 |
| cschwede | i’m wondering if we should add a sample EC policy that is suitable for SAIO. The default 10+4 doesn’t work on SAIO, thus people might raise a lot of questions on this | 05:51 |
| clayg | notmyname: I think they were mostly "could be cleaned up later" | 05:51 |
| clayg | cschwede: the example in the saio docs works on saio's :D | 05:52 |
| notmyname | got it | 05:52 |
| kota_ | clayg: oh, I didn't notice. so interesting but I was wrong, sorry. | 05:52 |
| cschwede | clayg: yeah, but not the example in swift.conf, and i heard there are people not reading the fine manuals. but maybe i’m just too pessimistic on this | 05:53 |
| notmyname | bah! | 05:53 |
| clayg | cschwede: not sure - i kinda like the swift.conf sample to be more "practical" | 05:54 |
| clayg | cschwede: if you want to setup a saio that has 16 disks so you can run a 10+4 you totally should! | 05:54 |
| cschwede | clayg: I want that SAIO! right now! | 05:54 |
| clayg | notmyname: what'd you do!? | 05:54 |
| clayg | cschwede: vagrant-swift-all-in-one can sorta do it | 05:55 |
| notmyname | patch 169897 | 05:55 |
| patchbot | notmyname: https://review.openstack.org/#/c/169897/ | 05:55 |
| clayg | you can get a 10+2 pretty easy anyway | 05:55 |
| cschwede | clayg: you’re right, let’s keep it as it is | 05:55 |
| clayg | notmyname: so what? | 05:55 |
| notmyname | clayg: gerrit is rejecting me because I'm trying to push that one, which is closed | 05:55 |
| clayg | cschwede: i don't know if i'm *right* - but that's how it says now - and it's *defensible* - which is about as much as I strive for these days | 05:56 |
| clayg | notmyname: so don't do that? | 05:56 |
| notmyname | clayg: yeah, I gotta figure out how to not have it as the base of my chain. the rebase picked it up too, but it had landed already | 05:57 |
| clayg | notmyname: I don't get it - you should ignore me and just fix it - i'm not helping anyone | 05:57 |
| notmyname | ok, there! | 05:59 |
| notmyname | ok, everything is final | 05:59 |
| notmyname | https://review.openstack.org/#/q/status:open+project:openstack/swift+branch:feature/ec_review+topic:bp/swift-ec,n,z | 05:59 |
| cschwede | \o/ | 05:59 |
| clayg | and notmyname scores the last word! | 05:59 |
| notmyname | ;-) | 05:59 |
| notmyname | well, final-final will be the merge commit tomorrow | 06:00 |
| notmyname | so if all this is ok, then we add our +2s +As to these and then clayg pulls the cork ont he first one | 06:00 |
| torgomatic | so is it all pushed up now? | 06:00 |
| notmyname | yup | 06:00 |
| cschwede | notmyname: you saw my comment on the docs? https://review.openstack.org/#/c/169990/16..17/doc/source/overview_erasure_code.rst | 06:01 |
| notmyname | uh oh | 06:01 |
| notmyname | what got lost? | 06:01 |
| clayg | notmyname: sorry man I thought i added that in from you diff - but then I probably lost it again when I when cherry picked the most recent doc commit :\ | 06:02 |
| notmyname | ah ok | 06:02 |
| clayg | process fail | 06:02 |
| clayg | you think i'd be better at this after doing it for two weeks | 06:02 |
| notmyname | no worries | 06:03 |
| notmyname | that's the only thing I see on that | 06:03 |
| clayg | mattoliverau: you're going to do the next feature brach - it'll be way better for everyone involved (except you, it'll suck for you, but i'm acctually ok with that) | 06:03 |
| mattoliverau | clayg: lol, but your so good at it :P | 06:03 |
| notmyname | mattoliverau: he volunteered! | 06:03 |
| mattoliverau | that was not a volunteer... aww crap.. | 06:04 |
| cschwede | hopefully all big features have been implemented before the rotation reaches Europe… ;) | 06:04 |
| clayg | yeah yeah - but like bus factor - you'll see - you guys only think i've done a decent job because you don't realize how good mattoliverau would be at it | 06:04 |
| clayg | cschwede: lol | 06:04 |
| mattoliverau | pfft as if, clayg you did an awesome job, thank you | 06:04 |
| cschwede | +1000! | 06:05 |
| clayg | mattoliverau: np, thanks for all the help | 06:05 |
| clayg | ok ok enough of that then | 06:05 |
| notmyname | clayg: do you still have that diff? | 06:05 |
| clayg | sure! | 06:05 |
| * clayg gets the last word after all! | 06:05 | |
| notmyname | clayg: too late ;-) | 06:07 |
| clayg | notmyname: nice ;) | 06:07 |
| notmyname | ok, now we're done (I hope) | 06:08 |
| notmyname | cschwede: thanks. good catch | 06:08 |
| clayg | yeah that diff didn't apply at all - i was going to just fix by hand but you must have seen the same thing | 06:08 |
| notmyname | I though I had some nicer words, but I don't remember what they were | 06:08 |
| clayg | yeah we'll see what acoles_away thinks about patch 169987 in the am | 06:08 |
| patchbot | clayg: https://review.openstack.org/#/c/169987/ | 06:08 |
| clayg | notmyname: oh here's the diff -> https://gist.githubusercontent.com/notmyname/9fd0a28fa36b9fe9c1de/raw/efbee3ead4268c3c4f19c40af42f83296b7842d5/gistfile1.txt | 06:09 |
| notmyname | oh, that didn't have it anyway | 06:09 |
| clayg | notmyname: oh, ok well I think it's all good | 06:10 |
| clayg | notmyname: it got the beta part | 06:10 |
| notmyname | :-) | 06:10 |
| * notmyname is going through and adding +2 to everything | 06:11 | |
| mattoliverau | notmyname: I thought you only said: "The fragment index" | 06:11 |
| *** krtaylor has quit IRC | 06:11 | |
| notmyname | mattoliverau: yeah, I think I had changed it locally to be more eloquent | 06:11 |
| mattoliverau | lol, cause I was going to say :P | 06:11 |
| clayg | it's fine | 06:12 |
| clayg | living document people | 06:12 |
| notmyname | ok, anyone else want to put a +2 on the patches? | 06:13 |
| notmyname | looks like torgomatic already has | 06:13 |
| cschwede | i | 06:13 |
| torgomatic | working on it | 06:13 |
| torgomatic | I'm looking over diffs from patchset N-1 to N just to be sure | 06:13 |
| *** Guest___ has quit IRC | 06:13 | |
| notmyname | that's smart ;-) | 06:13 |
| mattoliverau | and I'm +Aing stuff.. cause I can | 06:13 |
| notmyname | ok, in a bit once everyone has had a chance to do so, we'll land it all so that jenkins can do it's thing to land it on feature/ec_review | 06:14 |
| notmyname | those of you for whom it isn't the middle of the night, please kick jenkins as necessary over the next few hours (ie recheck) | 06:14 |
| notmyname | that way, in about 9ish hours (ie my tomorrow morning) I'll do the single merge commit to master | 06:15 |
| notmyname | and queue up the rest of the stuff pending to master | 06:15 |
| clayg | notmyname: charz: ok - so tomorrow we're going to do some massive reconstructor tests right? I wanna load up a see what it looks like to rebuild a a few TB of fragment archives! | 06:15 |
| cschwede | notmyname: when do you pull the plug? i’m wondering if you need to add a -2 on the first patch and clayg remove his one afterwards, to ensure you can pull the plug | 06:15 |
| notmyname | with the hope of tomorrow afternoon (my time), we'll have a SHA on master that is good for the RC | 06:15 |
| notmyname | cschwede: no, clayg can remove the -2 at any time | 06:16 |
| clayg | oh right - cause worst that happens is we merge to ec_review - sweet - i'm down | 06:16 |
| notmyname | cschwede: that will let the ec_review patches land. the single patch to master will be tomorrow. | 06:16 |
| cschwede | notmyname: ah, ok, i thought you would do that finally | 06:16 |
| notmyname | still, please don't merge anything else to master until after ec lands tomorrow | 06:16 |
| *** ChanServ changes topic to "HARD freeze of master in effect | EC Merge plan: https://etherpad.openstack.org/p/ec_merge_plan | Review Dashboard: http://goo.gl/vysJqI | Summit scheduling: https://etherpad.openstack.org/p/liberty-swift-summit-topics | Logs: http://eavesdrop.openstack.org/irclogs/%23openstack-swift/" | 06:17 | |
| clayg | mattoliverau: might you consider letting acoles have the final +A on patch 169987 when he wakes up in a few? | 06:18 |
| patchbot | clayg: https://review.openstack.org/#/c/169987/ | 06:18 |
| mattoliverau | sure :) | 06:18 |
| torgomatic | alright, I've done the bookkeeping stuff | 06:19 |
| notmyname | thanks :-) | 06:19 |
| * torgomatic is out | 06:20 | |
| mattoliverau | Did we decide to ignore peluse's ETage/Etag comment in patch #170339 | 06:21 |
| notmyname | patch 170339 | 06:22 |
| patchbot | notmyname: https://review.openstack.org/#/c/170339/ | 06:22 |
| clayg | mattoliverau: yeah :\ | 06:22 |
| clayg | peluse and I didn't quite grok what he was getting at | 06:22 |
| *** krtaylor has joined #openstack-swift | 06:22 | |
| clayg | mattoliverau: I think he was pointing out that some unless information was getting transfered/stored in Etag - but it was a pre-existing and benign issue. | 06:24 |
| mattoliverau | sounded like he just wanted to add a try/except cause in some case there could be an Etag header rather then ETag.. but he said its benign to swift so.. moving on | 06:24 |
| mattoliverau | clayg: what you said :P | 06:24 |
| clayg | mattoliverau: yeah I think since the reader always looks at ETag it just doesn't matter - idk, maybe it was worth fixing in the ECDiskFile case? | 06:26 |
| mattoliverau | meh, peluse can fix it in the next release :) | 06:27 |
| charz | clayg: notmyname Cool! So I can do tests on community cluster with ec_review in my tmr morning. | 06:27 |
| mattoliverau | if it needs to be fixed | 06:27 |
| *** Gues_____ has joined #openstack-swift | 06:28 | |
| clayg | charz: will that effect the CI testing? | 06:28 |
| notmyname | charz: ya, should be pretty soon. most of the delay will be getting the openstack-ci system to churn through the patches and land them | 06:28 |
| clayg | charz: do we have any other gear we can deploy on in the lab? | 06:28 |
| notmyname | clayg: I'm thinking that we should temporarily turn off the public patches and do some ec testing on those. but we have some other boxes too | 06:29 |
| notmyname | clayg: I've also got to put in the 8TB drives into the qa cluster | 06:29 |
| charz | clayg: yeah, we're going to turn off the jobs of jenkins tempoary. | 06:30 |
| *** tab__ has joined #openstack-swift | 06:31 | |
| clayg | charz: ok wfm | 06:31 |
| clayg | charz: what about the supermicro gear? | 06:32 |
| charz | clayg: Maybe I can turn on the jobs in your day and off it in night. And I can work on it in my day. | 06:33 |
| clayg | charz: I'd love to get my hands dirty too if you're not scared about me getting all up in the lab (hint: you should totally be scared about me getting all up the lab) | 06:34 |
| clayg | charz: I'll check with mlanner in the am and see if he thinks there's any boxes ic an play with | 06:34 |
| clayg | there's a couple of nodes from the kinetic rack I could probably borrow - maybe I fould find some object nodes to throw in with them | 06:34 |
| charz | clayg: Yeah! but It(smcluster) lacks a power cable or something. | 06:35 |
| clayg | we also have the old cluster we had attached to staging that was all beat up - I bet I could mess with that up | 06:35 |
| clayg | charz: ok, cool - i'm sure we can find something to keep me busy | 06:35 |
| notmyname | if there's one thing we have, it's a lot of ghetto swift clusters ;-) | 06:35 |
| * clayg is easily entertained | 06:35 | |
| clayg | notmyname: hey they're getting better! :P | 06:36 |
| notmyname | oh, definitely! | 06:36 |
| clayg | peluse: cschwede: mattoliverau: you guys be hoarding any gear for the cause? | 06:36 |
| charz | clayg: LOL | 06:36 |
| mattoliverau | acoles_away: I've left patch 170339 (reconstructor) [or cschwede] and 169987 (Per policy diskfile) for you to review and give a final +A of. Cause you know that code better then I :) | 06:37 |
| patchbot | mattoliverau: https://review.openstack.org/#/c/170339/ | 06:37 |
| clayg | normally everyone is all like "let us know if you need any gear for testing!" - and i'm all like "nah, mostly we just need code" | 06:37 |
| notmyname | clayg: go +2/+A the docs patch https://review.openstack.org/#/c/169990/ | 06:38 |
| *** acoles_away is now known as acoles | 06:38 | |
| clayg | ACOLES!!!! | 06:38 |
| notmyname | it's acoles!! | 06:38 |
| clayg | i knew you coud'nt stay away long | 06:38 |
| acoles | good morning! | 06:38 |
| mattoliverau | well, I have rackspace public cloud at my finger tips :) | 06:38 |
| clayg | notmyname: and you said you weren't going to be up till midnight | 06:38 |
| acoles | so what's the story? | 06:39 |
| clayg | mattoliverau: and all the CBS volumes we could want? | 06:39 |
| clayg | mattoliverau: :D | 06:39 |
| clayg | acoles: ignore everyone - go look at patch 169987 | 06:39 |
| patchbot | clayg: https://review.openstack.org/#/c/169987/ | 06:39 |
| notmyname | clayg: I've still got 21 minutes, and I'm all fired up now | 06:39 |
| mattoliverau | hmm.. well if it's for work purposes.. hmm ;) | 06:39 |
| notmyname | clayg: I also said I expect to be up well past midnight ;-) | 06:40 |
| acoles | ok | 06:40 |
| notmyname | acoles: the tl;dr is that we think we're done. | 06:40 |
| clayg | notmyname: he can't hear you | 06:40 |
| clayg | he's one with the code now | 06:40 |
| mattoliverau | acoles or cschwede: I left patch 170339 for one of you to +A as well | 06:41 |
| patchbot | mattoliverau: https://review.openstack.org/#/c/170339/ | 06:41 |
| mattoliverau | ^^ reconstructor | 06:41 |
| acoles | wait, you changed it... | 06:41 |
| *** mmcardle has joined #openstack-swift | 06:41 | |
| mattoliverau | as I haven't reviewed that patch as much as the others | 06:41 |
| * acoles reaches for his -2 | 06:41 | |
| notmyname | heh | 06:41 |
| acoles | :P | 06:41 |
| clayg | http://vignette4.wikia.nocookie.net/matrix/images/8/88/Neo_stops_bullets_2.JPG/revision/latest?cb=20060831032053 <- acoles | 06:41 |
| mattoliverau | lol | 06:42 |
| notmyname | acoles: also, as a summary, everything is frag-index everywhere now | 06:42 |
| kota_ | lol | 06:42 |
| clayg | mattoliverau: aww man the reconstructor commit message is all typo'd up - peluse was channling the clayg a little to hard when he typed it up :\ | 06:42 |
| mattoliverau | NOOO!!! | 06:43 |
| notmyname | it isn't pretty, but it's fine | 06:44 |
| mattoliverau | One thing I like about my timezone.. midnight for SFO is 5PM my time.. perfect timing for me :P | 06:44 |
| acoles | clayg: i always hated that verify stuff | 06:46 |
| clayg | acoles: pthpthtphthththt I thought you wrote it!? | 06:46 |
| notmyname | looks like we'll have some jenkins baby sitting to do: https://jenkins04.openstack.org/job/gate-swift-python27/1536/console | 06:47 |
| notmyname | already a py27 test failed | 06:47 |
| acoles | clayg: well i morphed what was in legacy | 06:47 |
| cschwede | notmyname: that’s another unsorted dict it seems | 06:48 |
| cschwede | look at the strings, abc-123 vs. 123-abc | 06:48 |
| clayg | cschwede: sweet does that mean mattoliverau can re-write the reconstructor commit message after all!? | 06:49 |
| cschwede | oh, no, it’s a set | 06:49 |
| clayg | cschwede: oh :\ | 06:50 |
| clayg | what is with you guys linking to jenkins? | 06:50 |
| notmyname | tests are still running, so I don't know of anything thing to link to | 06:50 |
| clayg | cschwede: ok the items are set's but the suffix ordering is a string | 06:50 |
| cschwede | clayg: so it might be that there are more places that might fail from time to time | 06:51 |
| clayg | cschwede: oh god that's gross :\ | 06:51 |
| clayg | I think i'd prefer to just make the replicate requests always send sorted.keys() - then the test can trust the order | 06:52 |
| clayg | cschwede: might help -> https://gist.github.com/clayg/116c3abc6566485f8b67 | 06:53 |
| cschwede | clayg: nice! | 06:54 |
| cschwede | clayg: i apply this quickly locally and run tests, to see if it matches current test constraints | 06:54 |
| acoles | cschwede: clayg: https://www.youtube.com/watch?v=I8GvDLDYhNM sing along with jenkins | 06:55 |
| clayg | acoles: what did you do? I'm randomly getting a KeyError on the del self.metadata['ETag'] line?! | 06:56 |
| clayg | test_reconstruct_fa_with_mixed_new_etag | 06:56 |
| *** nshaikh has joined #openstack-swift | 06:56 | |
| acoles | clayg: so about the verify problem - were you running tests with fst-post? how did you get a .meta file | 06:57 |
| acoles | clayg: what did *I* do? | 06:57 |
| clayg | acoles: yeah I was running with fast post | 06:57 |
| clayg | acoles: I didn't realize peluse added all those POSTs into the reconstructor tests | 06:58 |
| acoles | clayg: oh that was ambitious | 06:58 |
| acoles | clayg: so am i +A this right? | 06:58 |
| clayg | acoles: yup | 06:58 |
| acoles | done | 06:59 |
| notmyname | acoles: thanks! | 06:59 |
| notmyname | what do we want to do with the tests? new patch? what do you see cschwede? | 06:59 |
| *** PurpleJack has joined #openstack-swift | 06:59 | |
| cschwede | notmyname: give me a minute, my saio is stuck+ | 07:00 |
| cschwede | notmyname: clayg: with clays diff all unit test pass locally | 07:01 |
| notmyname | cschwede: did they fail before the patch? | 07:01 |
| cschwede | notmyname: i’d say let’s apply https://gist.github.com/clayg/116c3abc6566485f8b67 | 07:01 |
| notmyname | for you | 07:01 |
| clayg | cschwede: yeah I like that one | 07:01 |
| cschwede | notmyname: not for me, but for jenkins (the errors you linked above) | 07:01 |
| notmyname | right | 07:01 |
| *** jamielennox is now known as jamielennox|away | 07:02 | |
| clayg | i'm staring at test_reconstruct_fa_with_mixed_new_etag trying to understand how it ever worked (second reconstrct_fa should always keyerror because it's reusing the metadata that ETag was already popped from? | 07:02 |
| acoles | clayg: yeah i just reproduced the key error by looping that test | 07:02 |
| acoles | staring too | 07:02 |
| mattoliverau | clayg: is it a mutible defaults thing (like in the link you sent kota)? something isn't clearing | 07:03 |
| *** PurpleJack has quit IRC | 07:04 | |
| clayg | mattoliverau: idk, something is screwy - oh wait.. torgomatic fixed the with_ec_default ring | 07:04 |
| clayg | mattoliverau: ok i have cschwede's fix and this one - i need a new commit message | 07:05 |
| mattoliverau | lol, ok, on ec recon right | 07:06 |
| notmyname | ok, I've got a "new" patch chain with the sorted() call | 07:06 |
| clayg | acoles: womm -> https://gist.github.com/clayg/c2ee5a252a042ce19497 | 07:06 |
| notmyname | haven't pushed | 07:06 |
| clayg | notmyname: I have the other fix too - and mattoliverau is going to fix our commit message | 07:06 |
| notmyname | as always ;-) | 07:06 |
| cschwede | clayg: that fixes the etag issue? | 07:07 |
| clayg | yup | 07:07 |
| notmyname | clayg: you'll push all of it then? | 07:07 |
| clayg | notmyname: i can | 07:07 |
| notmyname | should just affect reconstructor and docs | 07:07 |
| notmyname | the last 2 in the chain | 07:07 |
| clayg | cschwede: see how it was resuing the same metadata dict for the second call to reconstruct? | 07:07 |
| notmyname | yay mutable types! | 07:08 |
| clayg | cschwede: the reason it was working is because the inserting the random invalid etag in the response list was preventing it from even getting to that code | 07:08 |
| clayg | then it started showing up because torgomatic fixed the default test policy for ec to be 10+4 instead of the faky 4+2 | 07:08 |
| cschwede | ah! | 07:09 |
| clayg | so the new_index think didn't always stick the bs response in the list of responses that the test grabbed | 07:09 |
| clayg | so it would make it all the way to "i'm about to rebuild this!" only to blow up with the keyerror that we never saw because it never got that far | 07:09 |
| cschwede | clayg: your fix for the etag issue works for me, looped it 50 times and passed always (without the patch it failed from time to time) | 07:09 |
| clayg | acoles: see that - test coupling with the default policies! | 07:09 |
| acoles | clayg: but the test assert DiskFileErro which should be raised before the diskfile thing is constructed - so the Key Error revealed that DiskFileError wasn't being raised | 07:10 |
| clayg | yah i think the real fix was the self.policy.ec_nparity instead of the hard coded 2 - but wrapping the metadata in the dict to prevent bleeding the mutated state doesn't hurt | 07:10 |
| notmyname | +2 | 07:10 |
| clayg | acoles: yeah when the random placement of the invalid etag fell off the ec_ndata it would raise keyerror instead showing the diksfileerror wasn't getting raised | 07:10 |
| acoles | clayg: the DIskFileError is only raised when number of responses is < ec_ndata, so is it just that sometimes the frag you break isn't needed to make enough good responses | 07:10 |
| clayg | i fixed the metadata thing only to find the diskfile error wasn't getting raised because of torgomatic's fix to the fake diskfile | 07:11 |
| clayg | so it's good - this is better | 07:11 |
| clayg | acoles: yeah you got it | 07:11 |
| clayg | but the self.policy.ec_nparity fix in that diff will square it 100% - make sure the invalid etag is in the bit of responses we need | 07:12 |
| acoles | clayg: oic, yup | 07:12 |
| ahale | man, you guys are busy! | 07:13 |
| clayg | ahale: it's something | 07:13 |
| notmyname | ahale: trying to give you new toys to play with ;-) | 07:13 |
| ahale | :) | 07:13 |
| tab__ | portante: i haven't managed to reponse to you yesterday regarding the question on which database is to be best used with Swift. Yes I am interested if the sqlite is OK to use in larger deployments ~1PB or is better to used MySQL/MariaDB? What's the practice? | 07:14 |
| clayg | acoles: I don't really see where i'm limiting the available responses to ndata tho... maybe there's still an unlikely failure in there | 07:14 |
| clayg | I want to say that on the 6+2 fake policy we were in this situation that a rebuild + 1 failure was as much erased codes as we could tolerate - but if I really did eat the 10+4 change we should be able to sustain one bad response | 07:15 |
| clayg | yeah with_ec_default is 10+4 | 07:15 |
| notmyname | tab__: there isn't a central DB in swift. and yes, there are many cluster that are _well_ above 1PB. (nor is there an opportunity to use anything other than sqlite in swift) | 07:15 |
| notmyname | tab__: https://swiftstack.com/openstack-swift/ is a good starting place to understand the architecture of how it works | 07:17 |
| mattoliverau | clayg: how about something like: http://paste.openstack.org/show/203823/ | 07:17 |
| acoles | clayg: it womm (your recon test change) | 07:17 |
| mattoliverau | thats my first attempt at rewording what was already there in the commit message | 07:18 |
| acoles | torgomatic: thanks for your reply on change log | 07:18 |
| mattoliverau | Capitalise the 'it' tho | 07:18 |
| notmyname | mattoliverau: looks good | 07:19 |
| clayg | mattoliverau: yeah roger - thanks | 07:19 |
| mattoliverau | close enough anyway | 07:19 |
| mattoliverau | probably needs to break up the long sentences but meh | 07:19 |
| clayg | acoles: although i feel like I should describe the purge/delete_reverted_objs a little more clearly than "it removes the partition" | 07:20 |
| clayg | notmyname: oh shit I don't have... no it's ok I got it | 07:20 |
| notmyname | :-) | 07:20 |
| notmyname | it's all in gerrit. you got it :-) | 07:20 |
| notmyname | clayg: let us know when you push those last 2 patches | 07:21 |
| tab__ | notmyname: ok thx. i am installing than mysql just for the purposes of keystone user authentication , but swift will run also sqlite for its own needs. | 07:27 |
| *** chlong has quit IRC | 07:27 | |
| clayg | ok - got the recon test fixup with the new commit message | 07:29 |
| *** jistr has joined #openstack-swift | 07:29 | |
| clayg | any other nits? | 07:29 |
| clayg | might be our last chance? | 07:29 |
| clayg | at some point we're going to have to let this baby be born | 07:29 |
| notmyname | I'm good | 07:29 |
| clayg | acoles: cschwede: mattoliverau: | 07:29 |
| mattoliverau | clayg: give birth man, push :P | 07:31 |
| clayg | ok all the git review sha's look right - so I think i rebased correctly | 07:31 |
| clayg | shit | 07:32 |
| notmyname | oh! | 07:32 |
| clayg | notmyname: how do you fix the thing where it thinks you're pushing up to a closed chnage? | 07:32 |
| notmyname | you gotta do `git review feature/ec_review` | 07:32 |
| clayg | I think i got a bogus changeid in there | 07:33 |
| notmyname | an extra one? or because of the closed change? | 07:33 |
| clayg | sigh, i'm not sure | 07:33 |
| *** krykowski has joined #openstack-swift | 07:33 | |
| notmyname | which one gave you an error? it's on the output message of git review | 07:34 |
| clayg | oh wait ohkay what happened now | 07:34 |
| clayg | did the think that makes git review smart like get reverted? | 07:34 |
| notmyname | not quite | 07:34 |
| notmyname | it's the rebase to master that lost it | 07:34 |
| notmyname | do you still have that in your patch chain locally? | 07:35 |
| clayg | notmyname: i got it - you were right | 07:35 |
| clayg | review feature/ec_review was the answer | 07:36 |
| notmyname | great! | 07:36 |
| notmyname | and I see the patches | 07:36 |
| clayg | https://review.openstack.org/#/c/170339/ is ready for jenkins worst! | 07:36 |
| clayg | and I think it's only that one and docs that need their +A added back | 07:36 |
| notmyname | interesting. it looks like gerrit kept the existing +2s. I haven't seen that before | 07:37 |
| notmyname | oh, nm | 07:37 |
| notmyname | makes sense, because I'm looking at the docs patch and it didn't effectively change. so gerrit kept it | 07:37 |
| clayg | "trivial rebase detected" | 07:37 |
| cschwede | iirc that happens if gerrit is able to detect a rebase | 07:37 |
| clayg | ok i think i'm about to turn in | 07:38 |
| mattoliverau | +A'ed the reconstructor.. go jenkins go! | 07:38 |
| clayg | i like pre-approving our diffs on the fly | 07:39 |
| clayg | it's like some sort of evolved form a pair programming | 07:39 |
| notmyname | :-) | 07:39 |
| clayg | global programming? | 07:39 |
| clayg | distributed hack-a-thon-style-mergefest? | 07:39 |
| *** ppai has quit IRC | 07:40 | |
| notmyname | clayg: thanks for everything! get some rest :-) | 07:40 |
| mattoliverau | lol, can we have that on a t-shirt :P | 07:40 |
| acoles | good work everyone! | 07:40 |
| clayg | well now I get to read zackmdavis's erausre code implimentation post | 07:40 |
| mattoliverau | At swift we do distributed hack-a-thon-style-mergefest-ifying | 07:41 |
| clayg | I really hope this makes it onto our blog - zackmdavis is a great author | 07:41 |
| * clayg desperatly wants to give away the ending but doesn't want to spoil it | 07:41 | |
| notmyname | wat | 07:43 |
| notmyname | https://jenkins03.openstack.org/job/gate-swift-python27/1358/console | 07:43 |
| cschwede | yay, first patch 169985 got merged | 07:43 |
| patchbot | cschwede: https://review.openstack.org/#/c/169985/ | 07:43 |
| cschwede | notmyname: more sorted stuff things :/ | 07:44 |
| notmyname | ya | 07:44 |
| clayg | lol! | 07:45 |
| kota_ | great! | 07:45 |
| notmyname | what https://review.openstack.org/#/c/169987/10 | 07:45 |
| clayg | i wonder if jenkins has a like a magic python build that tanslates dict.keys to sorted(dict.keys, reverse=True) to make sure no one sneaks in these little timebombs | 07:45 |
| notmyname | requires a local merge to resolve? | 07:45 |
| clayg | oh goodness - what did i do!? | 07:46 |
| notmyname | ugh. what's going on here | 07:47 |
| clayg | ok, i think i have the last of those suffix order comparision fixes in | 07:48 |
| cschwede | the error in https://jenkins03.openstack.org/job/gate-swift-python27/1358/console is strange, because it is sorted! | 07:48 |
| clayg | cschwede: well that's the ssync calls - I only added the sort to the replicate request | 07:48 |
| cschwede | clayg: but the test itself has a sorted() for both args just a few lines above | 07:49 |
| clayg | but that's just sorting the request tuples - not the lists in the tuples :\ | 07:49 |
| clayg | notmyname: ok - so what's the sha I need to rebase on? | 07:50 |
| clayg | a70782933 ec-test is merged - so that's probably a good one right? | 07:50 |
| cschwede | clayg: ah, yes, you’re right about that sorting of the lists | 07:51 |
| notmyname | clayg: no | 07:51 |
| notmyname | clayg: rebase on gerrit/feature/ec_review | 07:51 |
| clayg | notmyname: wow - i had wqa all typed in vim and everything | 07:51 |
| *** proteusguy has joined #openstack-swift | 07:51 | |
| *** ppai has joined #openstack-swift | 07:52 | |
| notmyname | rebasing on top of the now-current gerrit/feature/ec_review will sort out the merge commits and make it all cleaner | 07:52 |
| notmyname | also, side benefit, you get the `git review` sanity back | 07:52 |
| notmyname | and since the other merge commits have landed (2 of them, but we only needed one), it's also good with master | 07:52 |
| notmyname | short version: rebase on gerrit/feature/ec_review, and everything is good | 07:53 |
| clayg | hrmm.. ok but I think this will clear the +2's unless gerrit is whicked smart | 07:53 |
| notmyname | and should be trivial rebase for everything, except fro what you just patched | 07:53 |
| clayg | whoa - nice one gerrit! | 07:53 |
| notmyname | you just pushed it? | 07:54 |
| clayg | notmyname: yeah but I don't know about the .gitreview change coming back? that's good? | 07:54 |
| notmyname | yes, it's fine | 07:54 |
| notmyname | because it's still on that branch | 07:54 |
| clayg | ohhhhkay - if you say so | 07:54 |
| *** PurpleJack has joined #openstack-swift | 07:55 | |
| clayg | acoles: you wanna try again on patch 169987 before notmyname beats you to it? | 07:55 |
| patchbot | clayg: https://review.openstack.org/#/c/169987/ | 07:55 |
| acoles | k | 07:56 |
| acoles | clayg: +A'd | 07:56 |
| notmyname | great! | 07:56 |
| notmyname | I added (back) all of the +As ont he others | 07:57 |
| cschwede | +2/A on all patches - yeah! | 07:57 |
| cschwede | https://review.openstack.org/#/q/status:open+project:openstack/swift+branch:feature/ec_review,n,z | 07:57 |
| notmyname | yup, so a few minutes to make sure zuul likes it | 07:58 |
| acoles | clayg: i stared at your recon test some more, it works because the 'bad' response has newer timestamp so its etag is always preferred even though > ndata responses are received. there is only 1 response that has the preferred etag. | 07:58 |
| acoles | clayg: i expect you knew that! | 07:59 |
| notmyname | ok, all tests now running in the gate | 07:59 |
| mattoliverau | clayg: this is for you: http://oliver.net.au/memes/swift2.jpg (NOTE: I stole the image from google images, don't know if I'm allowed to use it) :P | 08:00 |
| notmyname | nice :-) | 08:00 |
| kota_ | mattoliverau: how great picture is :P | 08:01 |
| notmyname | sweet! py27 tests passed on the docs patch! | 08:04 |
| notmyname | that means it all worked | 08:04 |
| notmyname | (and that us USA-ians can go to bed) | 08:04 |
| mattoliverau | yay, thanks USA-ians! | 08:04 |
| notmyname | cschwede: mattoliverau: acoles: can you check zuul from time to time and recheck as necessary to get stuff landed? | 08:05 |
| cschwede | notmyname: sure! | 08:05 |
| mattoliverau | hai | 08:05 |
| notmyname | thanks! | 08:05 |
| acoles | yup all i'm going to do is watch zuul | 08:05 |
| clayg | acoles: if you manager catches you drinking at work today tell him clayg says it's fine | 08:06 |
| notmyname | I was about to say that if acoles watches zuul with a drink in hand, it's well earned and totally deserved :-) | 08:07 |
| acoles | heh | 08:07 |
| acoles | i'm not planning on going into the office | 08:08 |
| clayg | we are of one globally-distributed-hackathon-mergefest-ifying mind | 08:08 |
| notmyname | good! | 08:08 |
| clayg | night all - hoo ray ec - let's not do this again sometime! | 08:08 |
| notmyname | ok, I've got a big day tomorrow. final merge stuff, plus I'm doing an openstack podcast where I'll totally be bragging on everyone | 08:08 |
| notmyname | good night! | 08:08 |
| acoles | clayg: well done man! | 08:09 |
| clayg | notmyname: thanks champ! | 08:09 |
| cschwede | Good night folks, rest well! | 08:09 |
| clayg | acoles: all you brother | 08:09 |
| clayg | night | 08:09 |
| kota_ | have a good night USA-ians! | 08:09 |
| mattoliverau | night notmyname and clayg, you both rock! | 08:11 |
| mattoliverau | as does the rest of you, but keep working.. typey typey :P | 08:12 |
| mattoliverau | Well I'm going to go think about and cook dinner | 08:19 |
| *** jordanP has joined #openstack-swift | 08:20 | |
| *** ujjain has joined #openstack-swift | 08:20 | |
| cschwede | all patches passed the check queue, now in the gate :) | 08:45 |
| kota_ | cschwede: great :) | 08:46 |
| mattoliverau | yay! I was just about to say that too :) | 08:46 |
| acoles | who is getting the fireworks ready? | 08:47 |
| *** ujjain has quit IRC | 08:49 | |
| *** ujjain has joined #openstack-swift | 08:50 | |
| *** ppai has quit IRC | 08:51 | |
| *** aix has quit IRC | 08:53 | |
| *** ujjain has quit IRC | 08:57 | |
| ho | I'm ready :-) (Im back from meeting now. Amazing progress!!!) | 08:58 |
| *** km_ has joined #openstack-swift | 09:04 | |
| *** ppai has joined #openstack-swift | 09:05 | |
| *** km has quit IRC | 09:06 | |
| *** ppai has quit IRC | 09:14 | |
| *** bus-104_ has quit IRC | 09:15 | |
| acoles | heh. today i increased the scrollback lines option in my irc client :) that was a busy 12 hours! | 09:18 |
| acoles | clayg: peluse hats off to you guys for tracking down the .data + .meta assertion issue, sorry for that, i think way back peluse and i discussed our mutual dislike of those assertions. | 09:19 |
| *** aix has joined #openstack-swift | 09:20 | |
| acoles | i slavishly followed what was there but i do question the sense of an ondisk 'contract' is a system that expects failures, like one day that contract will break. and for you that day was yesterday. | 09:21 |
| acoles | ho: goo evening! | 09:21 |
| acoles | 'good' even | 09:22 |
| *** ujjain has joined #openstack-swift | 09:23 | |
| clayg | acoles: we need a "things we hate about diskfile" list so that in the liberty refactors we can make sure we get rid off all of them | 09:23 |
| ho | acoles: goo"d" morning :-) | 09:27 |
| *** ppai has joined #openstack-swift | 09:27 | |
| acoles | clayg: yup. | 09:28 |
| acoles | clayg: sleep!! per-policy-diskfiles just merged | 09:28 |
| * cschwede thinks clayg has a twin brother | 09:28 | |
| acoles | cschwede: the machine known as clayg could pass the turing test ;) | 09:30 |
| ho | patch 169989 was passed :-) | 09:31 |
| patchbot | ho: https://review.openstack.org/#/c/169989/ | 09:31 |
| cschwede | acoles: yes, definetely! | 09:31 |
| *** ppai has quit IRC | 09:34 | |
| cschwede | just a few seconds to go in the gate… | 09:36 |
| cschwede | PARTY TIME! | 09:40 |
| acoles | yay! | 09:40 |
| ho | fireworks: http://japan-fireworks.com/guide/miyajima.gif | 09:40 |
| kota_ | yey!! | 09:41 |
| acoles | i miss those days when jenkins would routinely throw us some random failures :) | 09:41 |
| ho | WOW!! | 09:41 |
| acoles | ho: nice! | 09:41 |
| acoles | http://www.londonlifestyleawards.com/london/wp-content/uploads/2011/11/Screen-shot-2011-11-03-at-12.52.011.png | 09:42 |
| ho | acoles: WONDERFUL!!! | 09:43 |
| acoles | cschwede: http://s2.germany.travel/media/gcb_1/mice/kongresse/staedte_1/hamburg_6/EVE_04_q_019_RET_1024x768.jpg | 09:44 |
| *** ppai has joined #openstack-swift | 09:46 | |
| mattoliverau | yay! | 09:48 |
| cschwede | acoles: awesome, that’s 10 minues by bike from home :D and I’m sure I have some photos of the Queen Mary 2 from that evening (the ship in the upper left corner) | 09:49 |
| *** silor has quit IRC | 09:50 | |
| ho | I'm gerous. If I could contribute swift community more, I could get great sense of achievement like yours. Next time :-) | 09:52 |
| kota_ | It was great time to see the ec_review branch finally landed so I'm going to leave my office to go back home :D | 09:55 |
| kota_ | much appreciated for the awsome works on swift community :) | 09:55 |
| *** ondergetekende_ has joined #openstack-swift | 10:02 | |
| *** proteusguy has quit IRC | 10:02 | |
| *** bkopilov has quit IRC | 10:02 | |
| *** ondergetekende has quit IRC | 10:02 | |
| *** kota_ has quit IRC | 10:03 | |
| openstackgerrit | Lorcan Browne proposed openstack/swift: Add lowest option to swift-recon disk usage output https://review.openstack.org/167236 | 10:55 |
| openstackgerrit | Christian Schwede proposed openstack/swift: More tests for swift recon https://review.openstack.org/173266 | 11:00 |
| *** ppai has quit IRC | 11:01 | |
| *** silor has joined #openstack-swift | 11:01 | |
| openstackgerrit | Christian Schwede proposed openstack/swift: WIP - More tests for swift recon https://review.openstack.org/173266 | 11:03 |
| acoles | cschwede: yes - more is good ^^ :) | 11:10 |
| *** proteusguy has joined #openstack-swift | 11:18 | |
| *** ujjain has quit IRC | 11:27 | |
| *** aix has quit IRC | 11:32 | |
| *** aix has joined #openstack-swift | 11:33 | |
| cschwede | yes, definetely. actually i was surprised about the low coverage (but then it’s only printing stuff, thus not that critical) | 11:40 |
| *** Anticimex has quit IRC | 11:41 | |
| ho | see you tomorrow! | 11:45 |
| acoles | ho: good night | 11:46 |
| *** ho has quit IRC | 11:47 | |
| *** jistr is now known as jistr|class | 11:47 | |
| *** km_ has quit IRC | 12:01 | |
| *** Gues_____ has quit IRC | 12:32 | |
| *** MVenesio has joined #openstack-swift | 12:42 | |
| *** mitz has quit IRC | 13:01 | |
| *** ozialien has joined #openstack-swift | 13:02 | |
| *** jistr|class is now known as jistr | 13:03 | |
| *** mitz has joined #openstack-swift | 13:04 | |
| peluse | hell, maybe I should go back to bed, yes! | 13:04 |
| *** jkugel1 has joined #openstack-swift | 13:06 | |
| *** ozialien_ has joined #openstack-swift | 13:12 | |
| *** ozialien has quit IRC | 13:12 | |
| *** ozialien_ is now known as ozialien | 13:12 | |
| *** dosaboy has quit IRC | 13:14 | |
| *** dencaval has joined #openstack-swift | 13:14 | |
| *** dosaboy has joined #openstack-swift | 13:14 | |
| notmyname | good morning | 13:17 |
| notmyname | git st | 13:18 |
| notmyname | that's a pretty picture :-) | 13:18 |
| *** rdaly2 has joined #openstack-swift | 13:21 | |
| *** haomaiwang has joined #openstack-swift | 13:21 | |
| *** fanyaohong has joined #openstack-swift | 13:21 | |
| *** ozialien has quit IRC | 13:23 | |
| notmyname | as far as patches pending for master goes, most things look good | 13:25 |
| notmyname | cschwede: can you rebase your patch on top of lorcan's current version? https://review.openstack.org/#/c/172752/ | 13:25 |
| openstackgerrit | Christian Schwede proposed openstack/swift: Add test for swift_recon.disk_usage https://review.openstack.org/172752 | 13:26 |
| cschwede | notmyname: done :) | 13:26 |
| notmyname | thanks | 13:26 |
| notmyname | cschwede: the dependent one (https://review.openstack.org/#/c/173266/) will eventually need it too, but isnt' queued to land yet | 13:27 |
| cschwede | notmyname: sure, but i need to fix it first | 13:28 |
| notmyname | heh | 13:28 |
| cschwede | i’m wondering why it passed locally but fails on jenkins. anyways, will do that later | 13:28 |
| notmyname | acoles: cschwede: thanks for checking on things during my (short) night. since everything looks good, I'lm gonna get myself ready | 13:28 |
| notmyname | today is button clicking in jenkins day to get everything landed :-) | 13:29 |
| cschwede | notmyname: great, can’t wait to see it land on master | 13:29 |
| notmyname | I'll start on that in about 1.5-2 hours | 13:29 |
| * notmyname goes to look for a shower and breakfast and coffee | 13:29 | |
| peluse | morning | 13:32 |
| *** jrichli has joined #openstack-swift | 13:33 | |
| *** jrichli has quit IRC | 13:40 | |
| peluse | FYI according to my IDE anyway: | 13:50 |
| peluse | ec_review | 13:50 |
| peluse | Code + docs: | 13:50 |
| peluse | LOC: 36872 | 13:50 |
| peluse | Total Lines: 84213 | 13:50 |
| peluse | Test Code: | 13:50 |
| peluse | LOC: 80534 | 13:50 |
| peluse | Total Lines: 97094 | 13:50 |
| peluse | master | 13:50 |
| peluse | Code + docs: | 13:50 |
| peluse | LOC: 34056 | 13:50 |
| peluse | Total Lines: 80564 | 13:50 |
| peluse | Test Code: | 13:50 |
| peluse | LOC: 72608 | 13:51 |
| peluse | Total Lines: 88289 | 13:51 |
| *** lpabon has joined #openstack-swift | 13:51 | |
| *** Gu_______ has joined #openstack-swift | 13:54 | |
| *** ppai has joined #openstack-swift | 13:56 | |
| *** Fin1te has joined #openstack-swift | 13:56 | |
| *** rdaly2 has quit IRC | 13:58 | |
| notmyname | ~2k sloc vs 8k tests? sounds about right | 14:00 |
| peluse | yeah, a rough 2K though that's for sure! | 14:01 |
| * notmyname packs up to get on the bus | 14:01 | |
| * peluse heads out to take daughter to school | 14:02 | |
| *** esker has joined #openstack-swift | 14:08 | |
| esker | Hoping the gracious folks here in the Swift IRC channel can help me out w/ a question I didn't know how to answer: does Swift "support" SMR drives (specifically, drive managed, host managed and host aware versions)? My understanding is that Swift really only needs a extended attributes file systems to sit atop and that drive type / characteristics are abstracted to it... but perhaps my understanding is 1) incorrect / outdated and/or 2) th | 14:11 |
| esker | ere's something fundamentally different about SMR drives? | 14:11 |
| esker | Sorry about the wall 'o text... | 14:11 |
| openstackgerrit | Christian Schwede proposed openstack/swift: More tests for swift recon https://review.openstack.org/173266 | 14:12 |
| *** Anticimex has joined #openstack-swift | 14:14 | |
| *** ppai has quit IRC | 14:21 | |
| *** Fin1te has quit IRC | 14:33 | |
| *** noye has joined #openstack-swift | 14:34 | |
| noye | what's recommended way to upgrading an existing openstack implemtation ? currently we only use swift and ceilometer | 14:34 |
| peluse | esker, I don't know much about SMR drives, someone else might, but yes wrt the need for a ex attr file system. Swift doesn't look any "deeper" than that :) | 14:40 |
| *** bkopilov has joined #openstack-swift | 14:43 | |
| *** zaitcev has joined #openstack-swift | 14:49 | |
| *** ChanServ sets mode: +v zaitcev | 14:49 | |
| *** acoles is now known as acoles_away | 14:49 | |
| esker | peluse: thanks! | 14:50 |
| *** acoles_away is now known as acoles | 14:51 | |
| *** Fin1te has joined #openstack-swift | 14:51 | |
| *** vinsh_ has joined #openstack-swift | 14:53 | |
| *** vinsh has quit IRC | 14:53 | |
| *** jrichli has joined #openstack-swift | 14:53 | |
| notmyname | hello | 14:54 |
| acoles | peluse: big day! some of us piled in late on but you have been driving the EC thing for months - so well done, great work! | 14:58 |
| esker | notmyname: quick question for you wrt SMR drives... is there anything special in Swift dealing w/ them? Is there anything that could be exploited given the varying types ( drive managed, host managed and host aware versions)? Or do they remain abstract from Swift given intermediating ex attr file system? | 14:59 |
| notmyname | esker: give me a few minutes.... | 14:59 |
| notmyname | (and I got your email | 14:59 |
| esker | notmyname: thanks! | 14:59 |
| notmyname | fun* time | 15:00 |
| openstackgerrit | John Dickinson proposed openstack/swift: Prevent unauthorized delete in versioned container https://review.openstack.org/173361 | 15:00 |
| notmyname | CVE-2015-1856 | 15:00 |
| *** ozialien has joined #openstack-swift | 15:01 | |
| *** Gu_______ has quit IRC | 15:01 | |
| notmyname | backports at https://review.openstack.org/#/c/173366/ and https://review.openstack.org/#/c/173363/ | 15:02 |
| notmyname | master at https://review.openstack.org/#/c/173361/ | 15:03 |
| notmyname | esker: ok, now that that's out of the way ;-) | 15:03 |
| notmyname | esker: short term, no. nothing special for the drive-managed SMR drives (yet). I'm getting some in my lab to test to empirically validate that | 15:04 |
| notmyname | esker: but in general (or conceptually) drive-managed SMR drives are fine | 15:04 |
| *** G________ has joined #openstack-swift | 15:04 | |
| notmyname | esker: longer-term, host-aware SMR drives will need some sort of filesystem that knows smr. so if that's on them, it's still ok. there's also a small question if swift itself can be smr aware. I don't know the answer to that yet | 15:05 |
| esker | notmyname: greatly appreciated! | 15:06 |
| esker | notmyname: let us know if we can help out on the host-aware investigation | 15:06 |
| * notmyname unflags your email ;-) | 15:06 | |
| esker | notmyname: delete if for that matter ;-) | 15:07 |
| *** pberis has quit IRC | 15:10 | |
| * notmyname prepares the ec to master merge commit | 15:12 | |
| *** jordanP has quit IRC | 15:13 | |
| peluse | acoles, thanks! by the level of contribution nobody would guess that you came in "late". In fact, I can't even type that with a straight face - man you really made some diving catches!! | 15:13 |
| *** ozialien has quit IRC | 15:14 | |
| acoles | peluse: it was fun. but I am enjoying a 'slow' day today :D | 15:14 |
| peluse | yeah, I hope clayg takes a few days off! Holy cow | 15:15 |
| notmyname | peluse: this looks interesting http://www.hgst.com/press-room/press-releases/HGST-ships-NVM-Express-Compliant-Solid-State-Drives-with-Industry-Leading-Performance-for-Cloud-and-Enterprise-Applications | 15:15 |
| notmyname | peluse: no kidding! | 15:15 |
| peluse | clayg's performance reminds me of when I was young :) | 15:15 |
| acoles | peluse: +1 i can't even remember being young | 15:16 |
| peluse | ha! | 15:16 |
| peluse | notmyname, I wonder how those stack up to ours... will dig up a link shortly... | 15:16 |
| notmyname | running all tests on the merge commit. then will post it | 15:17 |
| peluse | BTW the industry standard driver they mention (the windows one) is what myself and 4 others worked on developed over a little over a year just before I jumped in on this Swift stuff! | 15:17 |
| notmyname | cool | 15:17 |
| * notmyname is totally offended that the HGST press release doesn't credit peluse for actually allowing them to make the thing | 15:18 | |
| peluse | me too! :) | 15:18 |
| peluse | Here's ours: http://www.intel.com/content/www/us/en/solid-state-drives/solid-state-drives-dc-s3700-series.html | 15:18 |
| notmyname | why did I get unit test errors!!!???!!!?? | 15:19 |
| peluse | hmm, our random IOPS look better (well, benchmarks) | 15:19 |
| peluse | what!?! | 15:19 |
| notmyname | looking... | 15:20 |
| peluse | on running mine now | 15:20 |
| notmyname | ah ok | 15:20 |
| notmyname | peluse: ok, I need a small bit of help | 15:21 |
| peluse | ready | 15:21 |
| notmyname | peluse: the cve patch has a test that uses proxy_server.ObjectController() | 15:21 |
| notmyname | which isn't a thing any more | 15:21 |
| notmyname | so do I use ObjectControllerRouter? | 15:22 |
| peluse | 220, 221... whatever it takes | 15:22 |
| peluse | yes, I believe so but I need to look as well. my tests are almost done running | 15:22 |
| peluse | OK, what's the cve patch? my tests run OK | 15:23 |
| notmyname | it's linked above | 15:23 |
| notmyname | ok, I think I got it from a different test | 15:24 |
| notmyname | running tests again | 15:24 |
| peluse | OK, looking | 15:24 |
| peluse | cool, let me know. will look at the same time though | 15:24 |
| acoles | notmyname: yes router[policy] will return you the correct controller for the policy i believe | 15:25 |
| notmyname | acoles: peluse: after my local tests finish, will the 2 of you be around to +2/+A the merge commit? | 15:25 |
| peluse | yup | 15:26 |
| notmyname | hmm.. I see an "E" | 15:26 |
| peluse | notmyname, I don't see a patch for that going on the ec review branch, which of the 3 patches are you working on? | 15:26 |
| acoles | notmyname: yes | 15:26 |
| notmyname | peluse: I've got a merge commit from ec_review on top of https://review.openstack.org/#/c/173361/ | 15:27 |
| notmyname | bah! 'module' object has no attribute 'ReplicatedObjectController' | 15:27 |
| peluse | ahhh | 15:28 |
| notmyname | why? | 15:28 |
| notmyname | it's exactly the same as the test above | 15:28 |
| *** esker has quit IRC | 15:28 | |
| notmyname | it's failing in the test_denied_DELETE_of_versioned_object test in test proxy server | 15:28 |
| notmyname | ah! | 15:29 |
| acoles | notmyname: yeah i was about to say look at the test above controller = ReplicatedObjectController( | 15:29 |
| notmyname | yeah, I've got proxy_server.Replicated.... | 15:29 |
| acoles | notmyname: also see line 2364 self.app.obj_controller_router = proxy_server.ObjectControllerRouter() | 15:29 |
| notmyname | right. I got that | 15:30 |
| notmyname | I had replaced proxy_server.ObjectController with proxy_server.ReplicatedObjectController | 15:30 |
| notmyname | instead of just ReplicatedObjectController | 15:30 |
| notmyname | ok, I had already kicked of functests. waiting... | 15:30 |
| *** krykowski has quit IRC | 15:31 | |
| acoles | notmyname: you sorted? one of those 'helpful' imports | 15:31 |
| acoles | actually its not so bad | 15:31 |
| notmyname | acoles: the merge conflict? yes | 15:33 |
| acoles | yup, good | 15:33 |
| notmyname | functests pass | 15:33 |
| notmyname | rerunning unit tests | 15:33 |
| peluse | cool, just made the change here and started unit tests also | 15:34 |
| peluse | well, 2 changes | 15:34 |
| notmyname | looks like https://review.openstack.org/#/c/173363/ has a py26 failing gate tests | 15:35 |
| notmyname | I guess that's still checked on stable/juno | 15:36 |
| acoles | looking | 15:36 |
| *** itlinux_ has joined #openstack-swift | 15:36 | |
| itlinux_ | hello all | 15:36 |
| notmyname | unit tests passed for the marge commit | 15:36 |
| notmyname | probetests | 15:37 |
| notmyname | ... | 15:37 |
| itlinux_ | quick question what happen when the ring does not have enough amount on the node / partition will that split into two? | 15:37 |
| itlinux_ | thanks | 15:37 |
| *** pberis has joined #openstack-swift | 15:39 | |
| *** fanyaohong has quit IRC | 15:39 | |
| clayg | itlinux_: well you have replicas * 2 ** part-power partitions - that's normally enough to spread around | 15:39 |
| acoles | AssertionError: 'HTTP/1.1 5' != 'HTTP/1.1 2' on patch 173363 | 15:40 |
| patchbot | acoles: https://review.openstack.org/#/c/173363/ | 15:40 |
| itlinux_ | k thanks | 15:40 |
| notmyname | it's clayg! | 15:40 |
| peluse | whoa | 15:40 |
| *** vinsh_ is now known as Vinsh | 15:40 | |
| clayg | but if you have a LOT of nodes - the ring isn't going to make up any more partitions - so no - not split in half - you'd just get some nodes that would take more parts but don't get them | 15:40 |
| peluse | sleepwalking maybe? | 15:40 |
| itlinux_ | so it splits like inodes do in linux now | 15:40 |
| *** petertr7 has joined #openstack-swift | 15:40 | |
| clayg | peluse: no eating breakfest with the kids - mmmmm coffee | 15:41 |
| clayg | they're telling me i "type too fast" | 15:41 |
| itlinux_ | so the power of 2 is fine, but if the file is big it will be segmented into those partittion which the ring knows and will be allocated | 15:42 |
| notmyname | heh :-) | 15:42 |
| *** gyee has joined #openstack-swift | 15:42 | |
| clayg | yeah default max object size is like - 2 GiB? 5? | 15:42 |
| itlinux_ | hey notmyname :) I know you.. | 15:43 |
| clayg | so you can have a bunch of those in a part, and still have a bunch of parts on a disk | 15:43 |
| itlinux_ | 5 g | 15:43 |
| peluse | notmyname, FYI my stuf is a little behind yours but merge on top of CVE patch apssed all my units and is running probe now as well (just for addtl sanity check) | 15:43 |
| notmyname | peluse: great thanks. as soon as I see passing probetests here, i'll push it up | 15:44 |
| clayg | notmyname: did you ever come up with a mount/umount invocation that worked on your saio? | 15:44 |
| notmyname | clayg: yes | 15:44 |
| peluse | do share | 15:44 |
| clayg | notmyname: did you get it included in the tests? | 15:44 |
| * clayg bets he just used symlinks | 15:44 | |
| *** SkyRocknRoll has quit IRC | 15:44 | |
| peluse | def kill_drive(self, device): | 15:45 |
| peluse | if os.path.ismount(device): | 15:45 |
| peluse | os.system('sudo umount %s' % device) | 15:45 |
| peluse | else: | 15:45 |
| peluse | renamer(device, device + "X") | 15:45 |
| peluse | looks like he did :) | 15:45 |
| clayg | yeah i guess he fixed the unmount/umount thing | 15:45 |
| clayg | and then STOPPED TALKING TO US | 15:46 |
| notmyname | sortof that fix | 15:46 |
| notmyname | there were some other calls to kill_drive to use | 15:46 |
| notmyname | sorry, alpha_ori just walked in and was giving him the status update | 15:46 |
| peluse | all quiet on the western front... | 15:47 |
| notmyname | yes, everything is good | 15:47 |
| clayg | oh it's sorta later than I realized - I need to get will dressed | 15:47 |
| notmyname | and probetests just passed! | 15:47 |
| clayg | ok - i'll be in later - good news! good luck! | 15:47 |
| peluse | sweet, mine are still cooking | 15:47 |
| *** nshaikh has quit IRC | 15:47 | |
| openstackgerrit | John Dickinson proposed openstack/swift: Merge EC feature into master https://review.openstack.org/173398 | 15:47 |
| notmyname | thar she be | 15:48 |
| peluse | rock and rolla | 15:48 |
| peluse | will run all my shit on that one now | 15:48 |
| notmyname | peluse: clayg: acoles: please +2/+A as you see fit | 15:49 |
| peluse | will do | 15:49 |
| acoles | notmyname: will do just looking at the py26 fail | 15:49 |
| peluse | after tests run of course :) | 15:49 |
| *** pberis has quit IRC | 15:49 | |
| notmyname | acoles: thanks | 15:50 |
| *** jistr has quit IRC | 15:51 | |
| *** Fin1te has quit IRC | 15:51 | |
| acoles | notmyname: will 'reverify' alone work? figure we may as well get jenkins working in parallel | 15:53 |
| acoles | or do we have to have a bug for verify | 15:53 |
| acoles | ? | 15:53 |
| notmyname | acoles: should. or just "recheck no bug" should work | 15:53 |
| notmyname | acoles: except we have to wait for it to finish before doing that | 15:54 |
| notmyname | ah it did. and you did | 15:54 |
| acoles | yeah no sign of activity yet though. i though recheck only triggered checks and didn't do the verify/merge | 15:55 |
| notmyname | acoles: no, it failed again | 15:55 |
| notmyname | https://jenkins04.openstack.org/job/gate-swift-python26/766/console | 15:55 |
| peluse | notmyname, hey man, you left me off of the co-author list on the merge to master - trying to send a signal? | 15:56 |
| notmyname | ah did !? | 15:57 |
| peluse | you bastard :) | 15:57 |
| notmyname | peluse: so sorry. I picked one of the others, and you were the primary author on it | 15:57 |
| * notmyname fixes | 15:57 | |
| acoles | fight! | 15:57 |
| *** krtaylor has quit IRC | 15:57 | |
| *** G________ has quit IRC | 15:57 | |
| * peluse heads to the gym to lift something heavy and then put it back down | 15:58 | |
| openstackgerrit | John Dickinson proposed openstack/swift: Merge EC feature into master https://review.openstack.org/173398 | 15:58 |
| notmyname | doen | 15:58 |
| notmyname | done* | 15:58 |
| peluse | heh, OK now I can +2 :) | 15:58 |
| peluse | tox and func ran fine. running probe now | 15:58 |
| *** PurpleJack has quit IRC | 16:00 | |
| *** SkyRocknRoll has joined #openstack-swift | 16:00 | |
| acoles | peluse: me too lets see who has fastest cpu | 16:00 |
| acoles | peluse: wait, who did you work for again? ;) | 16:01 |
| peluse | I think I had a head start though | 16:01 |
| peluse | just passed my first S test, where is yours at? | 16:01 |
| acoles | 6 done | 16:01 |
| peluse | tick-tock | 16:02 |
| acoles | all passed! | 16:02 |
| acoles | just kidding | 16:02 |
| peluse | wow, I thought maybe you were running yours in some HP cloud or something :) | 16:02 |
| acoles | these tests used to run so much quicker before we wrote more :? | 16:03 |
| acoles | S | 16:03 |
| peluse | done, passed with flying colors (if thats possible) | 16:03 |
| *** Guest___ has joined #openstack-swift | 16:04 | |
| * acoles is still chugging along | 16:04 | |
| acoles | should have bought an a*d | 16:04 |
| peluse | so that just means you get to +A it I guess! | 16:04 |
| acoles | peluse: i spotted a typo in a comment in the tests! | 16:05 |
| acoles | -2 | 16:05 |
| notmyname | acoles: -2? ;-) | 16:05 |
| peluse | ha! | 16:06 |
| acoles | actually i saw it last night but let it go | 16:06 |
| * peluse searches for flights to the UK.... | 16:06 | |
| acoles | its been bugging me ever since | 16:06 |
| acoles | couldn't sleep | 16:06 |
| notmyname | we leave those things so we can support the drive-by-summit-pass committers ;-) | 16:06 |
| acoles | notmyname: exactly | 16:06 |
| acoles | probe passed | 16:06 |
| peluse | sweet | 16:06 |
| acoles | now func, i'm going backwards you see | 16:07 |
| notmyname | ok, I gotta get set up for http://www.nextcast.net/openstackpod/ in 30 minutes | 16:07 |
| notmyname | can someone look at the py26 errors on the juno backport? https://review.openstack.org/#/c/173363/ | 16:07 |
| peluse | sure, I can if acoles wasn't already in the middle of it | 16:08 |
| peluse | acoles? | 16:08 |
| acoles | notmyname: so far i saw obj server 409's so hoping it was just a timing blip but you said the test failed again :? womm | 16:08 |
| acoles | i'll go back to it when i've +A'd the EC merge | 16:09 |
| *** krtaylor has joined #openstack-swift | 16:12 | |
| *** esker has joined #openstack-swift | 16:13 | |
| acoles | peluse: notmyname +A | 16:32 |
| notmyname | yay!! | 16:32 |
| peluse | hel yes! | 16:32 |
| peluse | hell | 16:32 |
| peluse | :) | 16:32 |
| acoles | feel bad not waiting for everyone to get their +2 on it but figured we want this merged asap | 16:33 |
| acoles | torgomatic: clag: mattoliverau ^^ et al | 16:34 |
| peluse | does this system allow for +2 after the fact I wonder :) | 16:35 |
| acoles | peluse: i think so at least until its merged maybe. | 16:35 |
| acoles | peluse: i know it accepts and respects a -2 after a +A | 16:36 |
| *** Guest___ has quit IRC | 16:38 | |
| notmyname | thanks. /me logs off of IRC for a bit to do the podcast | 16:39 |
| openstackgerrit | Merged openstack/swift: Prevent unauthorized delete in versioned container https://review.openstack.org/173361 | 16:41 |
| *** SkyRocknRoll has quit IRC | 16:44 | |
| *** jkremer has joined #openstack-swift | 17:04 | |
| *** thumpba has joined #openstack-swift | 17:06 | |
| *** zhill has joined #openstack-swift | 17:15 | |
| *** aix has quit IRC | 17:18 | |
| acoles | notmyname: patch 173363 (stable/juno) merged on 2nd try, i suspect a race in the failed test can cause a 409 conflict and we got unlucky. looped test on my machine and could not reproduce. | 17:36 |
| patchbot | acoles: https://review.openstack.org/#/c/173363/ | 17:36 |
| *** jkugel1 has quit IRC | 17:54 | |
| *** Fin1te has joined #openstack-swift | 18:00 | |
| openstackgerrit | Merged openstack/swift: Merge EC feature into master https://review.openstack.org/173398 | 18:02 |
| acoles | ^^ woohoo | 18:02 |
| dencaval | Congratulations guys! | 18:05 |
| *** jamielennox|away is now known as jamielennox | 18:05 | |
| torgomatic | Is it now time for celebratory libations? | 18:06 |
| torgomatic | Heh, I got the email that said EC merged. Now I'm getting the dozens of merge conflict notifications. | 18:08 |
| *** krtaylor has quit IRC | 18:12 | |
| notmyname | yay! | 18:15 |
| notmyname | I'm going to go meet my wife for lunch, then I'll start on the other stuff to master | 18:15 |
| *** itlinux_ has quit IRC | 18:16 | |
| *** silor has quit IRC | 18:18 | |
| *** Guest__ has joined #openstack-swift | 18:21 | |
| *** tellesnobrega has quit IRC | 18:25 | |
| *** Guest__ has quit IRC | 18:25 | |
| *** tellesnobrega has joined #openstack-swift | 18:25 | |
| *** Guest__ has joined #openstack-swift | 18:26 | |
| *** tellesnobrega_ has joined #openstack-swift | 18:27 | |
| *** tellesnobrega_ has quit IRC | 18:27 | |
| *** tellesnobrega has quit IRC | 18:29 | |
| *** tellesnobrega has joined #openstack-swift | 18:29 | |
| *** Guest__ has quit IRC | 18:31 | |
| *** jkremer has quit IRC | 18:32 | |
| *** Anticimex has quit IRC | 18:33 | |
| *** Anticimex has joined #openstack-swift | 18:33 | |
| openstackgerrit | Christian Schwede proposed openstack/swift: Check if REST API version is valid https://review.openstack.org/168509 | 18:40 |
| *** jkugel1 has joined #openstack-swift | 18:49 | |
| *** annegentle has joined #openstack-swift | 18:50 | |
| *** Fin1te has quit IRC | 18:52 | |
| *** Fin1te has joined #openstack-swift | 18:59 | |
| *** acoles is now known as acoles_away | 19:02 | |
| *** krtaylor has joined #openstack-swift | 19:04 | |
| *** hugespoon has left #openstack-swift | 19:19 | |
| peluse | wheeeeeee | 19:22 |
| *** aix has joined #openstack-swift | 19:23 | |
| *** welldannit has joined #openstack-swift | 19:32 | |
| *** dencaval has quit IRC | 19:40 | |
| *** Fin1te has quit IRC | 19:40 | |
| notmyname | thanks for the rebases! | 19:46 |
| *** elmo_ has joined #openstack-swift | 19:46 | |
| torgomatic | master is unfrozen, yes? | 19:47 |
| notmyname | torgomatic: no | 19:47 |
| torgomatic | oh | 19:48 |
| notmyname | or consider it "managed" | 19:48 |
| notmyname | let me queue up some stuff for landing. | 19:48 |
| torgomatic | sure | 19:48 |
| torgomatic | notmyname: I'm gonna go abandon anything on feature/ec; objections? | 19:48 |
| notmyname | torgomatic: no. as long as you propose multi-range to master ;-) | 19:49 |
| torgomatic | notmyname: yeah, give me a few :) | 19:49 |
| notmyname | IIRC that's the only important one there | 19:49 |
| *** elmo has quit IRC | 19:49 | |
| peluse | wow, so we sould start scouring/reviewing/approving as needed the "needs approval" section? | 19:51 |
| notmyname | peluse: I'm doing that now. please hold off for a bit | 19:51 |
| peluse | roger | 19:52 |
| *** ozialien has joined #openstack-swift | 19:52 | |
| notmyname | honestly, I'm just trying to land the stuff that already has 2 +2s. and validating that there isn't any conflict | 19:52 |
| notmyname | and all that needs to land by tonight | 19:53 |
| notmyname | so, TBH, go take a day off ;-) | 19:53 |
| peluse | heh... I'm taking next week off. Going to a Volbeat convert in Denver next Fri (heading up just for the show) so figued I'd call the whole week a holiday :) | 19:54 |
| notmyname | heh ok. good. :-) | 19:55 |
| peluse | just updated trello, there's still cleanup work to be done there so its not ready to be retired (the board) | 19:55 |
| peluse | do we just land the spec? | 19:55 |
| notmyname | peluse: ya, per clayg's suggestion, I think it would be great to write down (trello? wiki? etherpad?) a list of stuff that we want to refactor/change | 19:56 |
| notmyname | ah, good question on the spec | 19:56 |
| notmyname | peluse: are there any open patches to the spec that haven't landed? | 19:56 |
| notmyname | doesn't look like it, unless I'm missing something | 19:56 |
| peluse | notmyname, nope, the last update I made should be sufficient to land it as is. It has a disclaimer of code being correct - spec was used to guide development, etc., etc., and the user docs are all correct | 19:56 |
| notmyname | of course | 19:57 |
| notmyname | peluse: can you please submit a patch to move it to the "done" folder? | 19:57 |
| peluse | sure, how does that work? | 19:57 |
| notmyname | `git mv` ;-) | 19:58 |
| notmyname | peluse: I'm not sure if links will work if moved. depends on how they were referenced. but the idea is to move the whole ec spec to the "done" directory instead if the "in_progress" directory | 20:00 |
| notmyname | links = images | 20:00 |
| peluse | will give it a try | 20:02 |
| peluse | what can go wrong? | 20:02 |
| notmyname | peluse: I can do it later if you don't want to | 20:02 |
| peluse | nah, no sweat | 20:02 |
| notmyname | ok, I think I have a plan for all of these pending patches | 20:06 |
| notmyname | I've queued them up locally all rebased off of each other. no conflicts, and tests all pass for them | 20:06 |
| notmyname | so I'll resubmit all of them so they land quickly and without the merge commits (ie FF merges for everything). should make it all easier for getting everythign landed today | 20:07 |
| notmyname | any objections? | 20:07 |
| peluse | nop | 20:07 |
| peluse | nope | 20:07 |
| notmyname | I had already approved 2 when I realized that this might be a better way. So I've got 7 queued up | 20:10 |
| openstackgerrit | Samuel Merritt proposed openstack/swift: EC: support multiple ranges for GET requests https://review.openstack.org/173497 | 20:10 |
| *** G________ has joined #openstack-swift | 20:10 | |
| torgomatic | notmyname: ^^ | 20:10 |
| notmyname | :-) | 20:10 |
| torgomatic | you asked me to port it; now you get to review it ;) | 20:10 |
| torgomatic | 11 files changed, 1625 insertions(+), 306 deletions(-) | 20:12 |
| notmyname | oh is that how it works? doh! | 20:12 |
| *** ozialien has quit IRC | 20:12 | |
| torgomatic | AND EVERY LINE IS GOLDEN | 20:12 |
| notmyname | ok, incoming.... | 20:15 |
| notmyname | I'll approve each of these to land | 20:15 |
| openstackgerrit | John Dickinson proposed openstack/swift: Add test for swift_recon.disk_usage https://review.openstack.org/172752 | 20:16 |
| openstackgerrit | John Dickinson proposed openstack/swift: Add lowest option to swift-recon disk usage output https://review.openstack.org/167236 | 20:16 |
| openstackgerrit | John Dickinson proposed openstack/swift: Check if REST API version is valid https://review.openstack.org/168509 | 20:16 |
| openstackgerrit | John Dickinson proposed openstack/swift: Add additional func tests for TempURLs https://review.openstack.org/155985 | 20:16 |
| openstackgerrit | John Dickinson proposed openstack/swift: Support HTTP_X_SERVICE_IDENTITY_STATUS in keystoneauth https://review.openstack.org/156634 | 20:16 |
| openstackgerrit | John Dickinson proposed openstack/swift: Refactor the getting roles logic in _keystone_identity https://review.openstack.org/166696 | 20:16 |
| openstackgerrit | John Dickinson proposed openstack/swift: Fix broken test_policy_IO_override test https://review.openstack.org/171593 | 20:16 |
| openstackgerrit | John Dickinson proposed openstack/swift: Check if device name is valid when adding to the ring https://review.openstack.org/169231 | 20:16 |
| notmyname | ok then | 20:18 |
| notmyname | should all be queued up to go | 20:18 |
| notmyname | and, wow, that + cleanign out the ec branch makes the dashboard really clean!! | 20:19 |
| notmyname | the only one starred right now is the last one on that patch chain | 20:20 |
| openstackgerrit | paul luse proposed openstack/swift-specs: Move Ec spec to done folder https://review.openstack.org/173503 | 20:24 |
| openstackgerrit | John Dickinson proposed openstack/swift: 2.3.0 authors and changelog updates https://review.openstack.org/172573 | 20:26 |
| notmyname | ^^ that should be the final patch on the RC (and in swift 2.3.0). the authors/changelog updates. I updated it to include a new author form the above patches and to add info about enforcing the api version | 20:26 |
| notmyname | the only reason there would be other stuff on 2.3.0 is if we find major show-stopping bugs that require an RC2 | 20:27 |
| notmyname | please keep master frozen until the authors/changelog patch lands. that will keep things easier if anything goes wrong in the gate | 20:28 |
| *** ChanServ changes topic to "HARD freeze of master in effect | Review Dashboard: http://goo.gl/vysJqI | Summary Dashboard: http://goo.gl/2By1qv | Summit scheduling: https://etherpad.openstack.org/p/liberty-swift-summit-topics | Logs: http://eavesdrop.openstack.org/irclogs/%23openstack-swift/" | 20:29 | |
| *** proteusguy has quit IRC | 20:34 | |
| notmyname | peluse: thanks | 20:37 |
| notmyname | peluse: (approved) | 20:37 |
| peluse | ya, gracias | 20:38 |
| *** proteusguy has joined #openstack-swift | 20:38 | |
| notmyname | torgomatic: clayg: mattoliverau: if you have a moment, can you add a 2nd +2 to https://review.openstack.org/#/c/172573/ | 20:38 |
| notmyname | (if you like it) | 20:38 |
| openstackgerrit | Merged openstack/swift: Clarify the description of backward compatibility in Keystoneauth's docstring https://review.openstack.org/166681 | 20:44 |
| *** ozialien has joined #openstack-swift | 20:47 | |
| *** proteusguy has quit IRC | 20:47 | |
| *** proteusguy has joined #openstack-swift | 20:48 | |
| *** Fin1te has joined #openstack-swift | 20:57 | |
| *** lpabon has quit IRC | 20:58 | |
| *** G________ has quit IRC | 21:00 | |
| *** ozialien has left #openstack-swift | 21:01 | |
| *** zhill has quit IRC | 21:10 | |
| *** annegentle has quit IRC | 21:14 | |
| *** zhill has joined #openstack-swift | 21:18 | |
| openstackgerrit | Merged openstack/swift-specs: Move Ec spec to done folder https://review.openstack.org/173503 | 21:21 |
| notmyname | last patch in the chain has passed all of our tests. (unit/func/docs/pep8) | 21:30 |
| *** Gu_______ has joined #openstack-swift | 21:33 | |
| *** Fin1te has quit IRC | 21:37 | |
| *** esker has quit IRC | 21:47 | |
| *** jkugel1 has quit IRC | 21:49 | |
| *** annegentle has joined #openstack-swift | 21:51 | |
| *** mmcardle has quit IRC | 21:52 | |
| *** Gu_______ has quit IRC | 21:54 | |
| notmyname | everything has passed the check queue. now it's all in the gate queue. ETA 60 minutes until all landed upstream | 22:11 |
| *** annegentle has quit IRC | 22:12 | |
| mattoliverau | Morning | 22:15 |
| notmyname | mattoliverau: hi! | 22:15 |
| notmyname | mattoliverau: zuul says 42 minutes until everything is landed and we have an RC! | 22:15 |
| mattoliverau | notmyname: how'd the pod cast go this morning? | 22:15 |
| notmyname | mattoliverau: good | 22:16 |
| clayg | notmyname: yeah i want to see it | 22:16 |
| notmyname | link might be up. let me see | 22:16 |
| mattoliverau | Yay! Awesome! | 22:16 |
| clayg | mattoliverau: yeah it better be | 22:16 |
| notmyname | https://www.youtube.com/watch?v=1ORJKia1pQs | 22:16 |
| clayg | could someone else at or near master run -> for i in {1..10}; do nosetests swift/test/unit/container/test_replicator.py; if [ $? -ne 0 ]; then break; fi; done and let me know if you see test_rsync_failure fail? | 22:17 |
| notmyname | clayg: on it | 22:17 |
| clayg | notmyname: well good - cause i'm going to be watching this video | 22:17 |
| notmyname | :-) | 22:18 |
| mattoliverau | Cool I'll give it a watch | 22:18 |
| clayg | hey look it's Nikki! | 22:18 |
| clayg | oh, Niki :\ | 22:18 |
| notmyname | clayg: so ./.unittests works just fine. running your thing fails (every time) | 22:22 |
| clayg | *every* time - or like if you run it 10 times it will *eventually* fail | 22:23 |
| clayg | like sometimes it works? | 22:23 |
| clayg | ... anyway - ok - i have a fix | 22:23 |
| notmyname | first time it fails | 22:23 |
| clayg | hrm | 22:23 |
| clayg | notmyname: curl https://gist.githubusercontent.com/clayg/ba58bb41d71d74695652/raw/961d1059eaa3b17e7bd25f25d8b1395f9867d3b8/gistfile1.diff | git apply and return? | 22:24 |
| clayg | retry? | 22:24 |
| notmyname | clayg: everything works. can you submit the patch based on https://review.openstack.org/#/c/172573/ | 22:29 |
| clayg | yeah sure | 22:29 |
| clayg | or i can do it - i mean i was going to submit it - so I could do it there, but i'm happy ignore it and let you handle it | 22:30 |
| notmyname | clayg: no, you got it. submit it dependent on the changelog patch, and we'll get it in the RC | 22:30 |
| clayg | ok | 22:30 |
| openstackgerrit | Merged openstack/swift: Add additional func tests for TempURLs https://review.openstack.org/155985 | 22:31 |
| openstackgerrit | Clay Gerrard proposed openstack/swift: Add support for multiple container-reconciler https://review.openstack.org/103779 | 22:33 |
| openstackgerrit | Clay Gerrard proposed openstack/swift: fixup random test failure https://review.openstack.org/173579 | 22:33 |
| openstackgerrit | Clay Gerrard proposed openstack/swift: wip: Cluster assisted Storage Policy Migration https://review.openstack.org/173580 | 22:33 |
| openstackgerrit | Merged openstack/swift: Check if device name is valid when adding to the ring https://review.openstack.org/169231 | 22:37 |
| openstackgerrit | Merged openstack/swift: Support HTTP_X_SERVICE_IDENTITY_STATUS in keystoneauth https://review.openstack.org/156634 | 22:44 |
| openstackgerrit | Merged openstack/swift: Refactor the getting roles logic in _keystone_identity https://review.openstack.org/166696 | 22:44 |
| *** fanyaohong has joined #openstack-swift | 22:48 | |
| *** proteusguy has quit IRC | 22:50 | |
| notmyname | looks like gerrit got confused | 22:59 |
| * notmyname go grabs the clue bat | 23:00 | |
| openstackgerrit | John Dickinson proposed openstack/swift: 2.3.0 authors and changelog updates https://review.openstack.org/172573 | 23:01 |
| openstackgerrit | John Dickinson proposed openstack/swift: Add test for swift_recon.disk_usage https://review.openstack.org/172752 | 23:01 |
| openstackgerrit | John Dickinson proposed openstack/swift: Add lowest option to swift-recon disk usage output https://review.openstack.org/167236 | 23:01 |
| openstackgerrit | John Dickinson proposed openstack/swift: Check if REST API version is valid https://review.openstack.org/168509 | 23:01 |
| openstackgerrit | John Dickinson proposed openstack/swift: Fix broken test_policy_IO_override test https://review.openstack.org/171593 | 23:01 |
| openstackgerrit | John Dickinson proposed openstack/swift: fixup random test failure https://review.openstack.org/173579 | 23:06 |
| notmyname | ok, everything rebased and reapproved | 23:08 |
| *** proteusguy has joined #openstack-swift | 23:09 | |
| *** chlong has joined #openstack-swift | 23:10 | |
| notmyname | ok, so https://review.openstack.org/#/c/173579/ is the last targeted patch for RC1. whenever gerrit gets around to landing it all | 23:11 |
| clayg | notmyname: that was good | 23:18 |
| notmyname | glad you liked it :-) | 23:18 |
| mattoliverau | notmyname: nice work (re: podcast) | 23:28 |
| *** chlong has quit IRC | 23:43 | |
| openstackgerrit | Samuel Merritt proposed openstack/swift-specs: Spec for symlinks in Swift. https://review.openstack.org/173609 | 23:45 |
| *** chlong has joined #openstack-swift | 23:45 | |
| *** ho has joined #openstack-swift | 23:46 | |
| *** wshao has joined #openstack-swift | 23:53 | |
| wshao | hi, any ideas on "SSL3_GET_SERVER_CERTIFICATE:certificate verify failed" when I use https for proxy-server? | 23:54 |
| wshao | the certs are self-signed. swift list/stat etc command would fail with the above error. | 23:54 |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!
