| *** MVenesio has joined #openstack-swift | 00:02 | |
| *** annegentle has quit IRC | 00:04 | |
| *** thumpba_ has quit IRC | 00:16 | |
| *** openstackgerrit has quit IRC | 00:22 | |
| *** openstackgerrit has joined #openstack-swift | 00:22 | |
| *** kota_ has joined #openstack-swift | 00:24 | |
| kota_ | I'm now back (to office) | 00:25 |
|---|---|---|
| kota_ | notmyname: still here? | 00:26 |
| clayg | kota_: he just walked out I think | 00:26 |
| kota_ | notmyname: I'm just now realizing we might need patch 175325 to land to stable/kilo | 00:26 |
| patchbot | kota_: https://review.openstack.org/#/c/175325/ | 00:26 |
| kota_ | clayg: morning! and thanks to work patch 175299 | 00:27 |
| patchbot | kota_: https://review.openstack.org/#/c/175299/ | 00:27 |
| kota_ | notmyname, clayg: 175325 is for a validation of k, m (number of data fragments and parity fragments) | 00:27 |
| *** zhill has quit IRC | 00:28 | |
| kota_ | notmyname, clayg: PyECLib 1.0.7 still have an decoding issue when the number of parity fragments is equals to or greater than the number of data fragments. | 00:28 |
| kota_ | notmyname, clayg: i.e. k=10, m=11 | 00:29 |
| clayg | kota_: seems reasonable - but i'm skeptical of it as a requirement for rc2 - notmyname is trying to get all that locked down tonite | 00:29 |
| kota_ | clayg: yeah, sounds ok. I don't have strong opinion to land it to rc2. | 00:29 |
| kota_ | clayg: Thanks to look at :D | 00:31 |
| *** openstack has joined #openstack-swift | 00:36 | |
| *** tsg has quit IRC | 00:37 | |
| *** gyee has quit IRC | 00:46 | |
| *** vinsh has quit IRC | 01:03 | |
| *** theanalyst has quit IRC | 01:04 | |
| *** theanalyst has joined #openstack-swift | 01:07 | |
| notmyname | patch 175325 | 01:16 |
| patchbot | notmyname: https://review.openstack.org/#/c/175325/ | 01:16 |
| notmyname | kota_: clayg: you tell me on that one for RC2 | 01:17 |
| *** MVenesio has quit IRC | 01:19 | |
| *** _joel has joined #openstack-swift | 01:22 | |
| *** vinsh has joined #openstack-swift | 01:31 | |
| *** vinsh has quit IRC | 01:38 | |
| *** vinsh has joined #openstack-swift | 01:39 | |
| mattoliverau | notmyname: it seems we are all asking this one by one... so my turn.. EC lib version numbers get confusing.. so on PyPi pyeclib 1.0.7 isn't bundled with liberasurecode 1.0.7.. only 1.0.5. Is this going to be fixed in PyPi before the release, if it is, then I assume patch 175325 isn't needed in RC2.. if it isn't and only liberasurecode 1.0.5 is bundled and installed then it will be required (at least if we suspect that | 01:42 |
| mattoliverau | people will test EC with more parity fragments then data). In a beta that might just be an edge case, so _could_ be added later. | 01:42 |
| patchbot | mattoliverau: https://review.openstack.org/#/c/175325/ | 01:42 |
| *** Hongtao has joined #openstack-swift | 01:45 | |
| Hongtao | who? | 01:45 |
| Hongtao | who have debuged the openstack swift with pycharm? | 01:46 |
| Hongtao | I need help ,thank you... | 01:46 |
| notmyname | mattoliverau: what I'm told is that they are still bundled on pypi. | 01:47 |
| Hongtao | Is there anybody? | 01:47 |
| notmyname | mattoliverau: or at least from a "packaging" standpoint (for whatever that means in pypi) | 01:47 |
| notmyname | mattoliverau: and if you don't have liberasurecode, pyeclib from pypi will install it | 01:48 |
| notmyname | Hongtao: what are you trying to find? | 01:48 |
| Hongtao | To develop middleware in swift. | 01:48 |
| Hongtao | So I should know how to debug it. | 01:49 |
| mattoliverau | notmyname: yeah, but at the moment (when I install pyeclib).. and I don't have librasurecode installed.. it installs liberausecod 1.0.5 (so assume 1.0.5 if bundled) (unless I've screwed something up). If 1.0.7 was bundled then that would be better | 01:49 |
| *** vinsh has quit IRC | 01:49 | |
| notmyname | mattoliverau: that's not what happened on my machine earlier today. I got 1.0.7 | 01:50 |
| *** vinsh has joined #openstack-swift | 01:50 | |
| mattoliverau | notmyname: lol, ok, maybe I've broken something :P | 01:50 |
| mattoliverau | or maybe the Oz pypi mirrors are crap :P | 01:50 |
| mattoliverau | if so I'll go complain to richard jones (in my team) as he runs the cheeseshop (pypi). | 01:51 |
| notmyname | Hongtao: you've already read http://docs.openstack.org/developer/swift/development_middleware.html | 01:51 |
| Hongtao | Almost out of my mind.... | 01:51 |
| notmyname | Hongtao: how do you normally use PyCharm to debug python server processes? | 01:51 |
| Hongtao | yeah.I have read it. | 01:52 |
| * notmyname doesn't use pycharm | 01:52 | |
| Hongtao | I can debug .py file with pycharm.But it doesn't work on swift. | 01:53 |
| *** bkopilov has quit IRC | 01:53 | |
| mattoliverau | Hongtao: I use pycharm to debug swift.. so you can, but depends on how you are doing it, swift is a bunch of services, so you either need to have each service running in pycharm, or you need to have SAIO running locally and then run the service you want to debug in pycharm. | 01:53 |
| notmyname | mattoliverau to the rescue ;-) | 01:53 |
| notmyname | /dinner time | 01:54 |
| Hongtao | Really ? | 01:54 |
| Hongtao | Can you give me some advices? | 01:54 |
| mattoliverau | Hongtao: sure | 01:54 |
| *** vinsh has quit IRC | 01:54 | |
| mattoliverau | notmyname: pycharm, once working, is a great way to actually debug tests, as you can step through nicely ;) | 01:55 |
| Hongtao | I searched on the internet . and I also don't know the correct way to configure. | 01:56 |
| mattoliverau | Hongtao: so you can't just debug a .py file.. cause to debug nicely you need to run the service in question. So I set up a run/debug configuration to run the swift/bin/<script> in question. | 01:57 |
| kota_ | I'm back again, I had a meeting. | 01:58 |
| mattoliverau | Hongtao: so on my development desktop, I run a SAIO (swift all in one) cluster. | 01:59 |
| Hongtao | what should i do now? | 01:59 |
| mattoliverau | Hongtao: I then start all the SAIO service I am about to debug. And create a run configuration for it in mycharm. | 02:00 |
| Hongtao | I also run a SAIO. when I have changed some code in swift.I run the command "swift-init all restart" | 02:01 |
| mattoliverau | Hongtao: say I want to run the swift-proxy-server.. try and run swift/bin/swift-proxy-server in pycharm | 02:01 |
| Hongtao | Is there any documents? | 02:01 |
| mattoliverau | Hongtao: that will fail, as you need to start it with the patch to swift.conf | 02:01 |
| mattoliverau | Hongtao: but now that you have tried to run the proxyserver when you goto edit configurations swift-proxy-server will be there, so you can add the parameter. | 02:02 |
| kota_ | notmyname, mattoliverau: Once I was talking with clayg I agreed patch 175325 is not needed in RC2 because it is a corner case and we could maintain it at the master. | 02:03 |
| patchbot | kota_: https://review.openstack.org/#/c/175325/ | 02:03 |
| mattoliverau | next time you try and run swift-proxy-server (or debug) it sould work, just add your break points and away you go. | 02:03 |
| mattoliverau | Hongtao: ^^ | 02:03 |
| Hongtao | what parameters? | 02:03 |
| mattoliverau | Hongtao: it tells you in the error (failed to start), you just add the path to the proxy config file. | 02:04 |
| *** lcurtis has joined #openstack-swift | 02:05 | |
| mattoliverau | whiich should be /etc/swift/swift-proxy.conf | 02:06 |
| mattoliverau | if you have setup the SAIO | 02:06 |
| mattoliverau | Hongtao: ^ | 02:06 |
| mattoliverau | Hongtao: if you just used swift-init restart all, then the proxy will fail to start cause you started a proxy via swift-init, so stop the proxy service first (if your debugging proxy) | 02:07 |
| mattoliverau | Hongtao: swift-init proxy stop | 02:07 |
| mattoliverau | Hongtao: then run/debug in pycharm so it can bind to the right port. | 02:07 |
| Hongtao | oh | 02:09 |
| mattoliverau | Hongtao: depending on what you are debugging you don't need to start all daemons. I.E if you are just testing the object-server, then you can make a run configuration and run it, then hit it will curl or whatever. | 02:10 |
| Hongtao | If I want to debug the swift.and I should close the proxy and run swift-init proxy stop? | 02:10 |
| Hongtao | Oh i know .Thank you very much! | 02:11 |
| mattoliverau | Hongtao: so long sotry short, you need to setup a run configuration on the actual scripts in bin/ but swift is a cluster with a bunch on daemons, so depending on what you are debugging then SAIO or setting up a bunch of run configurations might be in order to get a working environment | 02:12 |
| mattoliverau | Hongtao: but running a unit test is as simple as rightclicking on the test method/class or file in pycharm and saying run. | 02:13 |
| *** spudly has joined #openstack-swift | 02:22 | |
| *** _joel has quit IRC | 02:34 | |
| *** _joel has joined #openstack-swift | 02:37 | |
| Hongtao | mattoliverau : can you write a document for us who don't know how to handle. | 02:39 |
| mattoliverau | Hongtao: when I have some free time, maybe I'll turn it into a blog post, so everyone can read it :) | 02:40 |
| *** vinsh has joined #openstack-swift | 02:40 | |
| Hongtao | Very very thank you! where can I Contact your | 02:42 |
| Hongtao | where can I contact you directly | 02:42 |
| mattoliverau | Hongtao: you can always reach me here in channel, or email me at matt@oliver.net.au if you email me, I'll ping you back once I've had a chance to blog it.. I really need to get back into blogging anyway :P | 02:44 |
| Hongtao | Ok thank you | 02:45 |
| openstackgerrit | Janie Richling proposed openstack/swift: WIP - working on the encryption feature. https://review.openstack.org/157907 | 02:47 |
| *** vinsh has quit IRC | 03:11 | |
| *** vinsh has joined #openstack-swift | 03:11 | |
| *** vinsh has quit IRC | 03:15 | |
| *** silor has joined #openstack-swift | 03:16 | |
| *** bkopilov has joined #openstack-swift | 03:19 | |
| *** vinsh has joined #openstack-swift | 03:32 | |
| *** vinsh has quit IRC | 03:32 | |
| *** vinsh has joined #openstack-swift | 03:32 | |
| *** vinsh has quit IRC | 03:37 | |
| openstackgerrit | Merged openstack/swift: Bump PyECLib version from 1.0.3 to 1.0.7 https://review.openstack.org/174225 | 03:47 |
| kota_ | great, PyECLib bump landed to master. | 03:50 |
| mattoliverau | woo | 04:00 |
| kota_ | mattoliverau: I wonder if you could put workflow +1 patch 175299? | 04:02 |
| patchbot | kota_: https://review.openstack.org/#/c/175299/ | 04:02 |
| mattoliverau | I can now that the recheck passed ;) | 04:03 |
| mattoliverau | kota_ done | 04:04 |
| kota_ | mattoliverau: thanks! :) | 04:05 |
| *** _joel has quit IRC | 04:09 | |
| openstackgerrit | Merged openstack/swift: Included step in development_saio.rst for installing dependencies in requirements.txt. https://review.openstack.org/175594 | 04:11 |
| openstackgerrit | Merged openstack/swift: SAIO instructions: ensure ~/bin exists before copying into it https://review.openstack.org/175613 | 04:20 |
| *** SkyRocknRoll has joined #openstack-swift | 04:46 | |
| *** SkyRocknRoll has joined #openstack-swift | 04:46 | |
| openstackgerrit | Merged openstack/swift: Use reconstruct insetad of decode/encode https://review.openstack.org/175299 | 04:48 |
| *** lcurtis has quit IRC | 04:52 | |
| *** root has joined #openstack-swift | 05:03 | |
| *** root is now known as Guest82049 | 05:04 | |
| *** tsg has joined #openstack-swift | 05:05 | |
| *** Guest82049 has quit IRC | 05:08 | |
| *** Hongtao has quit IRC | 05:08 | |
| *** ds_ has joined #openstack-swift | 05:09 | |
| *** ds_ has quit IRC | 05:11 | |
| *** hbhdytf has joined #openstack-swift | 05:11 | |
| *** rsFF has quit IRC | 05:12 | |
| *** rsFF has joined #openstack-swift | 05:12 | |
| *** zaitcev has quit IRC | 05:20 | |
| *** hbhdytf has quit IRC | 05:32 | |
| cschwede | Good Morning! | 05:55 |
| kota_ | cschwede: morning! | 05:59 |
| cschwede | kota_: Hello Kota! | 05:59 |
| *** proteusguy has joined #openstack-swift | 06:05 | |
| mattoliverau | cschwede: guten morgen! | 06:09 |
| cschwede | mattoliverau: Hello Matthew! | 06:10 |
| *** kota_ has quit IRC | 06:11 | |
| *** proteusguy has quit IRC | 06:14 | |
| *** krykowski has joined #openstack-swift | 06:29 | |
| *** jamielennox is now known as jamielennox|away | 06:34 | |
| *** pcaruana has quit IRC | 07:11 | |
| *** bill_az has quit IRC | 07:14 | |
| *** slo_ has joined #openstack-swift | 07:26 | |
| *** geaaru has joined #openstack-swift | 07:41 | |
| *** jamielennox|away is now known as jamielennox | 07:43 | |
| *** jistr has joined #openstack-swift | 07:47 | |
| *** foexle has joined #openstack-swift | 07:58 | |
| *** jordanP has joined #openstack-swift | 08:16 | |
| *** acoles_away is now known as acoles | 08:31 | |
| *** nottrobin has quit IRC | 08:57 | |
| *** nottrobin has joined #openstack-swift | 08:59 | |
| openstackgerrit | Romain LE DISEZ proposed openstack/swift: Support of the Linux socket option SO_REUSEPORT https://review.openstack.org/137659 | 09:04 |
| *** krykowski has quit IRC | 09:09 | |
| acoles | notmyname: just catching up - did we cut RC2? re revert-y patches 175076 and 174875, like clayg said in my scrollback, i am comfortable with those issues being addressed during liberty | 09:18 |
| acoles | clayg: yeah, i agree there is almost certainly going to be some case (e.g. a race creating diskfiles) when we will be glad to have multi frag indexes. | 09:22 |
| acoles | so nothing wasted | 09:23 |
| acoles | clayg: i'd like to see that star trek quote in a commit message :) | 09:25 |
| openstackgerrit | Prashanth Pai proposed openstack/swift: versioned writes middleware https://review.openstack.org/134347 | 10:01 |
| *** thurloat has quit IRC | 10:04 | |
| *** thurloat has joined #openstack-swift | 10:04 | |
| *** silor has quit IRC | 10:07 | |
| *** krykowski has joined #openstack-swift | 10:37 | |
| *** jistr has quit IRC | 11:02 | |
| *** jistr has joined #openstack-swift | 11:04 | |
| *** kei_yama has quit IRC | 11:09 | |
| *** silor has joined #openstack-swift | 11:14 | |
| *** km has quit IRC | 11:28 | |
| *** tsg has quit IRC | 11:30 | |
| *** MVenesio has joined #openstack-swift | 11:34 | |
| *** MVenesio has quit IRC | 11:56 | |
| *** bkopilov has quit IRC | 12:12 | |
| *** aix has joined #openstack-swift | 12:21 | |
| *** MVenesio has joined #openstack-swift | 12:35 | |
| *** jistr has quit IRC | 12:42 | |
| *** jistr has joined #openstack-swift | 12:43 | |
| *** SkyRocknRoll has quit IRC | 12:55 | |
| *** bill_az has joined #openstack-swift | 12:56 | |
| *** jistr is now known as jistr|mtg | 13:01 | |
| *** sandywalsh has quit IRC | 13:11 | |
| *** sandywalsh has joined #openstack-swift | 13:15 | |
| *** esker has joined #openstack-swift | 13:26 | |
| openstackgerrit | Alistair Coles proposed openstack/swift: Fix tempauth acl checks when simplejson has no speedups https://review.openstack.org/159530 | 13:37 |
| *** EmilienM has quit IRC | 13:40 | |
| *** EmilienM has joined #openstack-swift | 13:41 | |
| openstackgerrit | Alistair Coles proposed openstack/swift: Merge branch 'master' into feature/crypto https://review.openstack.org/175866 | 13:44 |
| *** imkarrer_ has joined #openstack-swift | 13:51 | |
| *** tsg has joined #openstack-swift | 13:56 | |
| *** Fin1te has joined #openstack-swift | 14:01 | |
| *** petertr7 has joined #openstack-swift | 14:05 | |
| *** jrichli has joined #openstack-swift | 14:06 | |
| *** lcurtis has joined #openstack-swift | 14:10 | |
| *** annegentle has joined #openstack-swift | 14:10 | |
| *** bkopilov has joined #openstack-swift | 14:12 | |
| *** jistr|mtg is now known as jistr | 14:26 | |
| *** vinsh has joined #openstack-swift | 14:26 | |
| *** fanyaohong has joined #openstack-swift | 14:28 | |
| *** annegentle has quit IRC | 14:28 | |
| *** annegentle has joined #openstack-swift | 14:28 | |
| *** vinsh_ has joined #openstack-swift | 14:29 | |
| *** vinsh has quit IRC | 14:31 | |
| *** wbhuber has joined #openstack-swift | 14:34 | |
| *** _joel has joined #openstack-swift | 14:39 | |
| *** _joel has left #openstack-swift | 14:41 | |
| *** Fin1te has quit IRC | 14:49 | |
| *** tsg has quit IRC | 14:52 | |
| *** sileht has quit IRC | 14:57 | |
| *** breitz has joined #openstack-swift | 14:58 | |
| *** vinsh_ is now known as Vinsh | 15:02 | |
| *** Fin1te has joined #openstack-swift | 15:10 | |
| *** krykowski has quit IRC | 15:14 | |
| *** BaconIsKing has joined #openstack-swift | 15:16 | |
| *** jigsaw has joined #openstack-swift | 15:16 | |
| jrichli | Hi everyone. Has there ever been talk about adding a "slowdown" tunable for the replicator process much | 15:19 |
| jrichli | like the updater has a slowdown feature? | 15:19 |
| *** breitz has quit IRC | 15:20 | |
| *** breitz has joined #openstack-swift | 15:20 | |
| openstackgerrit | Merged openstack/swift: Merge branch 'master' into feature/crypto https://review.openstack.org/175866 | 15:22 |
| *** MVenesio has quit IRC | 15:23 | |
| *** MVenesio has joined #openstack-swift | 15:24 | |
| *** thumpba has joined #openstack-swift | 15:25 | |
| *** david-lyle has quit IRC | 15:36 | |
| *** thumpba_ has joined #openstack-swift | 15:38 | |
| *** thumpba has quit IRC | 15:41 | |
| *** Fin1te has quit IRC | 15:48 | |
| *** gyee has joined #openstack-swift | 15:50 | |
| *** tsg has joined #openstack-swift | 15:54 | |
| *** Vinsh_ has joined #openstack-swift | 15:56 | |
| *** sileht has joined #openstack-swift | 15:57 | |
| *** BaconIsKing has quit IRC | 15:58 | |
| *** Vinsh has quit IRC | 15:59 | |
| *** david-lyle has joined #openstack-swift | 16:02 | |
| notmyname | good morning | 16:06 |
| notmyname | 9:06am and I've already had 2 meetings | 16:06 |
| lcurtis | Hey notmyname | 16:07 |
| notmyname | today I'll be working on getitng RC2 squared away. the goal is to have a commit SHA on stable/kilo by EoD so ttx can tag RC2 | 16:07 |
| *** lpabon has joined #openstack-swift | 16:17 | |
| hurricanerix | morning | 16:17 |
| lcurtis | all...would i be better off using ext4 over xfs for small objects (4k and under)? | 16:18 |
| *** lpabon has quit IRC | 16:21 | |
| *** jistr has quit IRC | 16:21 | |
| notmyname | lcurtis: normally the reason xfs is chosen is because it performs better when there are a lot of inodes on the device (eg as disks get full). | 16:22 |
| notmyname | hurricanerix: hello | 16:22 |
| lcurtis | Thanks notmyname | 16:24 |
| *** jordanP has quit IRC | 16:29 | |
| *** fanyaohong has quit IRC | 16:39 | |
| clayg | acoles: so you have the plan to combine the handoff revert fixes with the 409 ssync fixes and make everything amazing and rainbows!? | 16:41 |
| clayg | ujjain: morning! | 16:43 |
| acoles | clayg: right now i am drowning in merge conflicts on my fast-post reviews so a rainbow would be good :P | 16:44 |
| acoles | clayg: are you referring to my option 3b comment? | 16:44 |
| clayg | FAST-POST! liberty is going to be the best development cycle yet. | 16:44 |
| clayg | acoles: yeah | 16:44 |
| acoles | liberty!! i'm waiting for 'p'=='post' cycle! | 16:45 |
| clayg | acoles: i had some notion that the solution wasn't like neither a nor b - but rather some of a AND b | 16:45 |
| clayg | acoles: oh no, l m n o p - it's no far off i guess :P | 16:45 |
| notmyname | jrichli: sort of. and it's sort of there today with the run_pause/rsync options/concurrency options | 16:46 |
| clayg | jrichli: I think the auditor has some crazy rate limit turning - what does the updater do now? | 16:46 |
| notmyname | jrichli: but, no there's not anything like "only process this many parts per unit time" | 16:47 |
| acoles | m for 'maybe', n for 'nearly', 'o' for 'oh-so-close'... | 16:47 |
| *** annegentle has quit IRC | 16:47 | |
| acoles | clayg: just going back to read 3a and 3b again... | 16:47 |
| clayg | so slowdown is like a small sleep on each top level dir in the async folder | 16:48 |
| acoles | clayg: so my 3b was meant as a superset of 3 | 16:49 |
| acoles | otherwise it would be 4 :P | 16:49 |
| acoles | clayg: so fix the KeyError, take the 409 on the chin, make ssync tell us what *did* get sync'd, clean those up, no duplicates | 16:50 |
| clayg | acoles: nah it's take the one good idea I had (don't keyerror) and add in the important don't blow up ssync entirely on 409 from your fix | 16:51 |
| clayg | acoles: I think maybe (False, {hash=>ts_in_sync}) will capture most of want we want right? | 16:51 |
| clayg | acoles: you think there's any hope the reciever might be able to eat the 409 and continue syncing the rest of the objects? Is there anything in the protocol for the reciever to return individual errors on files? | 16:53 |
| *** Vinsh_ has quit IRC | 16:54 | |
| *** Vinsh has joined #openstack-swift | 16:54 | |
| acoles | clayg: I think we have to be careful about eating the 409 and not letting the sender know that we did, and no nothing in protocol now to return individual status :/ | 16:55 |
| acoles | clayg: sender must NOT delete frags that got 409 on receiver, so ideally receiver would send back individual status, or list of all successful sync'd hashes | 16:56 |
| acoles | but thats a protocol change | 16:56 |
| acoles | clayg: and i didn't build allowance for *that* protocol change into EC | 16:57 |
| *** portante_ is now known as portante | 16:59 | |
| clayg | acoles: ic, yeah that's a rub :\ | 16:59 |
| acoles | clayg: we get part way with (False, {hash=>ts_in_sync}) where the dict at least has all objs the rx did not want to sync | 17:00 |
| acoles | clayg: that allows cleanup of dups on seconf pass | 17:00 |
| acoles | second | 17:00 |
| clayg | acoles: so there is a space in the reciever's response to include *an* error - it could probably be updated with something like the missing check protocol to include per hash errors | 17:03 |
| clayg | but I don't quite follow - how a 409 becoming a failure would let us clean up stuff on the second pass? | 17:05 |
| clayg | like you could maybe cleanup everythign that synced *before* the 409, but if the 409 blows up everything how would we skip it and make progress? | 17:06 |
| clayg | yeah it's definately not in-sync, but their may be other stuff in the part that could sync | 17:07 |
| *** geaaru has quit IRC | 17:08 | |
| *** david-lyle has quit IRC | 17:08 | |
| acoles | clayg: its more complicated that i fisrst thought, ...the rx 409 don't cause the updates to stop until a threshold failure count is reached, | 17:08 |
| clayg | i might try to have reciever send back a <hash status> line for errors and let sender pop from the insync map | 17:08 |
| acoles | clayg: so some 409s will be tolerated, updates continue, but at end of updates rx will raise an error to sender so we always get False, {} back to reconstructor | 17:09 |
| acoles | ^^ on master that is | 17:09 |
| acoles | lots of 409s will cause updates to blow up early if i read it correctly | 17:10 |
| clayg | acoles: it's weird - saw that code - but i guess it's not really important unless you have lots of objects - the reciever will say "i blew up" on self.failures | 17:10 |
| acoles | so the thing about 2nd pass and dups is... | 17:10 |
| clayg | acoles: yeah - violently agree | 17:10 |
| acoles | any 409 will cause False,{} return from sender :( | 17:10 |
| acoles | so we never get to cleanup stuff that is in sync before updates even started | 17:11 |
| acoles | but it would be easy to return False, {stuff rx never even wanted} | 17:11 |
| clayg | acoles: oh interesting - so even tho revert doesn't do a pre-flight suffix check - it *does* do the missing check before building requests | 17:12 |
| acoles | clayg: yes (remember me whining about that being wasted work when suffixes are equal :P) | 17:13 |
| clayg | acoles: which part - the missing check is wasted effort - oh or the whole revert? | 17:14 |
| acoles | so at end of missing check we can populate can-delete-objs or whatever with objs in sync | 17:14 |
| acoles | the missing check | 17:14 |
| *** EP77 has joined #openstack-swift | 17:15 | |
| clayg | acoles: i'm not sure which part is wasted | 17:16 |
| clayg | on a part forklift like a rebalance it's pretty stupid | 17:16 |
| acoles | clayg: ok that comment was a red herring sorry! IF we had rechecked suffixes for reverts, and if they had matched, then calling ssync_sender would cause a wasted missing_check that would result in nothing being missing - but we didn't do that, i was just wandering off topic. | 17:17 |
| clayg | heheheh | 17:17 |
| acoles | so back on track... | 17:18 |
| clayg | yeah | 17:18 |
| clayg | acoles: so i'm going to play around with having ssync reciever return per hash errors - seems simple enough | 17:18 |
| clayg | or useful | 17:18 |
| clayg | or something like that | 17:18 |
| clayg | acoles: my only reservation is that ssync missing_check ... why are we even trying to send a object that's just going to 409!? | 17:20 |
| acoles | clayg: +1 like there is a 'phase of handshakes' there just begging for it : rx sends UPDATES START, UPDATES END but we could insert a list of hash->status in between those, or even just successful hashes, just like the MISSING CHECK responses | 17:20 |
| clayg | want = df.timestamp < timestamp | 17:20 |
| acoles | good point! | 17:21 |
| acoles | clayg: because in missing_check rx tries to open the diskfile with the frag-index-to-revert, which does not exist, but in updates it gets a 409 because another frag exists | 17:23 |
| acoles | clayg: but rx could just try to open *any* frag... | 17:23 |
| acoles | during missing_check, then we'd avoid the 409 | 17:24 |
| clayg | interesting - but we'd probably also delete it - since it never *wanted* it - it'd be considered in-sync :\ | 17:24 |
| acoles | oh yeah, oops :P | 17:25 |
| acoles | i have another idea brewing.... | 17:25 |
| clayg | acoles: but you're right about hte frag index thing | 17:25 |
| acoles | clayg: so for fast-post, i change missing check responses to include hash *and timestamps* of meta, content-type, so sender can figure out if it needs to do a PUT or a POST or both... | 17:27 |
| clayg | acoles: yeah I love that | 17:27 |
| acoles | clayg: so we add to the missing-check response a 'want-but-can't-accept' flag | 17:28 |
| acoles | clayg: or somehow represent that semantic | 17:28 |
| clayg | ok | 17:28 |
| clayg | my mind is a little blown | 17:28 |
| acoles | clayg: then sender knows (a) its not in sync but (b) don't bother sending | 17:28 |
| clayg | yes, i have to admit that does seem to be *exactly* what we need to solve the issue at hand | 17:29 |
| acoles | clayg: so i'm making this up as i go now... | 17:29 |
| clayg | acoles: you should call it quits already like an hour ago! | 17:29 |
| acoles | clayg: but maybe rx includes a list of frag indexes that it has along with timestamps it has | 17:29 |
| clayg | acoles: I'm glad I caught you tho | 17:29 |
| acoles | oh, i forgot, i have really exciting merge conflicts to resolve... :P | 17:30 |
| clayg | acoles: that *would* be very ec specific | 17:30 |
| clayg | acoles: it's a really situation to say - yeah I'd love to have that, but I can't hold it right now - can you just hang onto it for me | 17:30 |
| acoles | clayg: ummm, so it would,. we could call them node indexes or wotnot :P | 17:30 |
| clayg | i need to stew - you've given me a lot to chew on - thank you | 17:31 |
| acoles | clayg: ok, so maybe not return frag indexes, but there would be a way to say 'not in sync but don't send now' | 17:32 |
| acoles | clayg: sure, i'll stew some more too... | 17:32 |
| *** annegentle has joined #openstack-swift | 17:32 | |
| clayg | I'm heading in | 17:33 |
| clayg | acoles: thanks again - talk to you later | 17:33 |
| acoles | clayg; k i may be gone but around tomorrow | 17:34 |
| clayg | acoles: yeah of course - give me time to stew ;) | 17:34 |
| *** aix has quit IRC | 17:37 | |
| *** annegentle has quit IRC | 17:38 | |
| *** annegentle has joined #openstack-swift | 17:39 | |
| *** thumpba_ has quit IRC | 17:48 | |
| *** EP77 has left #openstack-swift | 17:50 | |
| *** EP77 has quit IRC | 17:50 | |
| *** annegentle has quit IRC | 17:53 | |
| *** jigsaw_ has joined #openstack-swift | 17:54 | |
| jrichli | notmyname: clayg: thanks for the info. When you talked of rsync options, do you mean the rsync bwlimit? Also, is there a reason that slowdown makes sense for updater, but not replicator? | 18:03 |
| *** annegentle has joined #openstack-swift | 18:05 | |
| jrichli | oh, well I guess the replicator has a sleep between run loops, while the updater has a sleep between processing objects | 18:08 |
| *** annegentle has quit IRC | 18:10 | |
| notmyname | jrichli: yeah, the rsync options in the object replicator config, but also configuring rsync itself (like the max_connections settion) | 18:14 |
| notmyname | *setting | 18:14 |
| jrichli | ok | 18:15 |
| notmyname | be right back | 18:16 |
| *** patchbot has quit IRC | 18:16 | |
| *** notmyname has quit IRC | 18:16 | |
| *** notmyname has joined #openstack-swift | 18:17 | |
| *** ChanServ sets mode: +v notmyname | 18:17 | |
| notmyname | did you miss me? | 18:17 |
| *** patchbot has joined #openstack-swift | 18:18 | |
| *** bsdkurt has quit IRC | 18:27 | |
| notmyname | rd | 18:30 |
| notmyname | I'm running a last sanity check on the 4 patches for RC2 | 18:31 |
| notmyname | then I'll push those and land them on stable/kilo | 18:31 |
| notmyname | anything else need to be in RC2? this is the chance to say something | 18:32 |
| clayg | back | 18:36 |
| notmyname | clayg: I'm finishing up rc2 stuff now | 18:36 |
| clayg | notmyname: well what do we have? | 18:36 |
| notmyname | the 4 listed on https://wiki.openstack.org/wiki/Swift/PriorityReviews | 18:36 |
| clayg | etag pop fix, pyeclib 1.0.7, reconstruct-hack-removal | 18:37 |
| notmyname | and a translation one | 18:37 |
| clayg | oh wow - super important | 18:37 |
| notmyname | :-) | 18:37 |
| *** openstackgerrit has quit IRC | 18:37 | |
| *** Fin1te has joined #openstack-swift | 18:37 | |
| *** openstackgerrit has joined #openstack-swift | 18:37 | |
| *** acoles is now known as acoles_away | 18:38 | |
| notmyname | I post-facto created the LP bugs for them for the rc2. and updated commit messages. as soon as tests finish running locally, I'll push them to gerrit | 18:38 |
| notmyname | and I'm thinking I'll jsut go ahead and land them | 18:38 |
| notmyname | clayg: any objections? | 18:38 |
| *** david-lyle has joined #openstack-swift | 18:39 | |
| clayg | notmyname: no i don't think so | 18:39 |
| notmyname | ok | 18:42 |
| *** jigsaw has quit IRC | 18:42 | |
| *** jigsaw_ has quit IRC | 18:43 | |
| *** david-ly_ has joined #openstack-swift | 18:46 | |
| *** david-lyle has quit IRC | 18:46 | |
| *** imkarrer__ has joined #openstack-swift | 18:50 | |
| *** imkarrer_ has quit IRC | 18:53 | |
| notmyname | ok, patches pushed to gerrit and into the gate queue | 18:53 |
| notmyname | https://review.openstack.org/#/q/status:open+project:openstack/swift+branch:stable/kilo+topic:rc2patches,n,z | 18:53 |
| *** silor1 has joined #openstack-swift | 19:03 | |
| *** silor has quit IRC | 19:05 | |
| *** slo___ has joined #openstack-swift | 19:12 | |
| *** david-ly_ has quit IRC | 19:18 | |
| *** BaconIsKing has joined #openstack-swift | 19:18 | |
| *** gyee has quit IRC | 19:23 | |
| *** Fin1te has quit IRC | 19:37 | |
| *** MVenesio has quit IRC | 20:04 | |
| *** MVenesio has joined #openstack-swift | 20:04 | |
| *** imkarrer__ has quit IRC | 20:10 | |
| *** silor1 has quit IRC | 20:10 | |
| *** imkarrer has joined #openstack-swift | 20:10 | |
| *** BaconIsKing has quit IRC | 20:14 | |
| *** david-lyle has joined #openstack-swift | 20:21 | |
| *** zaitcev has joined #openstack-swift | 20:31 | |
| *** ChanServ sets mode: +v zaitcev | 20:31 | |
| *** Gu_______ has joined #openstack-swift | 20:32 | |
| *** imkarrer has quit IRC | 20:39 | |
| *** imkarrer has joined #openstack-swift | 20:40 | |
| *** Fin1te has joined #openstack-swift | 20:43 | |
| *** openstackgerrit has quit IRC | 20:52 | |
| *** openstackgerrit has joined #openstack-swift | 20:52 | |
| *** foexle has quit IRC | 20:52 | |
| *** imkarrer__ has joined #openstack-swift | 20:52 | |
| *** Fin1te has quit IRC | 20:55 | |
| *** imkarrer has quit IRC | 20:56 | |
| *** haomaiwang has quit IRC | 21:07 | |
| *** haomaiwang has joined #openstack-swift | 21:07 | |
| clayg | notmyname: there's one call in the reconstrctor that uses node['ip instead of node['replication_ip | 21:14 |
| notmyname | clayg: gate seems wedged with some other issue, so we could land it on master and then in rc2 if we do it quickly | 21:15 |
| clayg | notmyname: thanks gate wedge! | 21:20 |
| *** haomaiwang has quit IRC | 21:35 | |
| *** haomaiwang has joined #openstack-swift | 21:36 | |
| openstackgerrit | Clay Gerrard proposed openstack/swift: Use node's replication_ip to GET fragments during rebuild https://review.openstack.org/176090 | 21:37 |
| notmyname | clayg: thanks for doing the LP bug!! | 21:38 |
| clayg | notmyname: i thought you'd like that ;) | 21:39 |
| *** haomaiwang has quit IRC | 21:40 | |
| *** haomaiwang has joined #openstack-swift | 21:42 | |
| notmyname | ah, this is the issue on the gate https://bugs.launchpad.net/openstack-gate/+bug/1446847 | 21:45 |
| openstack | Launchpad bug 1446847 in OpenStack-Gate "stable/juno gate is busted on stevedore/cliff releases on 4/21" [Undecided,New] | 21:45 |
| notmyname | fishbowl sessions will have something like 150-200+ seats. working sessions have around 30 | 21:56 |
| mattoliverau | Morning | 21:56 |
| notmyname | mattoliverau: hi | 21:57 |
| notmyname | mattoliverau: can you check out https://review.openstack.org/#/c/176090/ please? | 21:58 |
| mattoliverau | Yup, sure thing | 21:58 |
| torgomatic | so we have a session called something like "Erasure Coded Cryptography in Your Global Swift Cluster" in a fishbowl, and while all the code tourists are there, we go have a couple working sessions and get things done | 21:59 |
| torgomatic | notmyname: that's how it works, right? | 21:59 |
| notmyname | torgomatic: actually yes. and every working session will have the same title on the schedule | 21:59 |
| torgomatic | notmyname: nice! | 21:59 |
| notmyname | and we'll have an etherpad or something to map sessions to what we're covering | 21:59 |
| *** slo___ has quit IRC | 22:00 | |
| notmyname | (but you forgot "docker" in the fishbowl session) | 22:00 |
| * torgomatic wishes he could forget Docker | 22:00 | |
| *** gyee has joined #openstack-swift | 22:01 | |
| zaitcev | AttributeError: 'Input' object has no attribute 'send_hundred_continue_response' | 22:05 |
| zaitcev | Oh, brother | 22:05 |
| zaitcev | The pace of progress is taking its toll on my body. | 22:05 |
| notmyname | zaitcev: old eventlet? | 22:06 |
| zaitcev | notmyname: you bet | 22:07 |
| *** Vinsh has quit IRC | 22:08 | |
| *** imkarrer__ has quit IRC | 22:11 | |
| jrichli | 150-200+ seats => (*_*) | 22:15 |
| notmyname | jrichli: no worries | 22:16 |
| jrichli | easy for you to say. You are good at public speaking | 22:16 |
| jrichli | but I guess I will live through it :-) | 22:17 |
| zaitcev | A number of community colleges in the U.S. have open enrollment programs where you can take public speaking courses. Just FYI. I'm saying it as an immigrant who's not a native speaker. | 22:18 |
| jrichli | zaitcev: thanks. I have had practice, and I guess I do ok. I will just always get nervous, that's all! | 22:21 |
| notmyname | the community cluster failures are spurious. the networking is being redone on them, and they weren't silenced before that started. I've turned the jobs off, so shouldn't be getting errors any more | 22:27 |
| notmyname | mattoliverau: go ahead and +A. don't wait for jenkins. that way we know what needs to happen as soon as it's available | 22:28 |
| clayg | notmyname: so the gate has been let go? or you just fixed the glitch? | 22:28 |
| mattoliverau | Hai | 22:28 |
| notmyname | (and jenkins is a computer. no need to be nice and give "him" a break) | 22:28 |
| *** jrichli has quit IRC | 22:29 | |
| notmyname | clayg: the community qa cluster? I "fixed" it by disabling all 3 jobs | 22:29 |
| zaitcev | heh | 22:29 |
| clayg | so it'll just work itself out naturally - We always like to avoid confrontation, whenever possible - Problem is solved from your end. | 22:29 |
| notmyname | whenever mlanner gets done redoing the networking, I'll turn it on. or he will. or someone will. point is, it'll get turned back on | 22:30 |
| notmyname | ok, I'll cherry-pick that patch to rc2 | 22:31 |
| *** annegentle has joined #openstack-swift | 22:32 | |
| openstackgerrit | John Dickinson proposed openstack/swift: Use node's replication_ip to GET fragments during rebuild https://review.openstack.org/176110 | 22:33 |
| notmyname | (also, yay. the change to report on all of our branches now works) | 22:34 |
| clayg | heh | 22:34 |
| *** annegentle has quit IRC | 22:38 | |
| *** lcurtis has quit IRC | 22:38 | |
| zaitcev | I always save these e-mail reminders to review something interesting "when I get back from a store" and then I always find them merged already. | 22:40 |
| mlanner | sorry about that. i'm hurrying as fast as i can, trying to break as little as possible. | 22:40 |
| notmyname | mlanner: no worries at all :-) | 22:41 |
| notmyname | don't rush it. it's fine | 22:41 |
| notmyname | zaitcev: the obvious solution is that you no longer go to the store | 22:41 |
| *** wbhuber has quit IRC | 22:43 | |
| *** km has joined #openstack-swift | 22:44 | |
| *** haomaiwang has quit IRC | 22:46 | |
| *** annegentle has joined #openstack-swift | 22:46 | |
| *** haomaiwang has joined #openstack-swift | 22:47 | |
| *** haomaiwang has quit IRC | 22:49 | |
| *** haomaiwang has joined #openstack-swift | 22:50 | |
| *** petertr7 has quit IRC | 22:51 | |
| *** annegentle has quit IRC | 23:00 | |
| *** kota_ has joined #openstack-swift | 23:08 | |
| kota_ | good morning | 23:08 |
| notmyname | hello | 23:09 |
| notmyname | tsg: what's the ec helper tool? where can I find that? | 23:09 |
| *** Gu_______ has quit IRC | 23:13 | |
| clayg | the ec helper tool is a lie! | 23:14 |
| notmyname | well ok then! | 23:14 |
| notmyname | that answers that | 23:15 |
| tsg | notmyname: pyeclib_conf_tool.py is part of pyeclib | 23:15 |
| clayg | tsg: no it's not - i have pyeclib installed and I have no such script | 23:17 |
| notmyname | ditto | 23:17 |
| notmyname | but it's in the tools dir of the source | 23:17 |
| *** zhill has joined #openstack-swift | 23:18 | |
| tsg | notmyname: that's correct - included with source only at the moment | 23:18 |
| notmyname | tsg: I have no idea how to use this | 23:18 |
| notmyname | tsg: I've got a 5 node cluster. each with 12 drives. (60 total). what ec numbers do I use? | 23:19 |
| mattoliverau | kota_: morning | 23:19 |
| kota_ | mattoiverau: morning | 23:20 |
| tsg | notmyname: the only "guidance" it can provide today is around the available EC backends and how those backends would perform (bytes/second) | 23:20 |
| tsg | notmyname: keving was in the middle of adding support for providing better guidance around data durability - I will check with him on the status | 23:23 |
| notmyname | tsg: what does "perf rank #10" mean? it's the 10th fastest? it's the faster? does it have anything to do with speed? | 23:26 |
| kota_ | tsg: Once, I wrote a script to calcurate durability, here https://github.com/bloodeagle40234/DurabilitySimulator/blob/master/simulator.py | 23:28 |
| tsg | notmyname: rank0 is the fastest - this is compute performance only though | 23:29 |
| kota_ | tsg: but it is needed to update the description for users, yet :P | 23:29 |
| tsg | notmyname, kota_: let me see keving can shed some light | 23:29 |
| notmyname | tsg: mostly I'm starting with my question (5 servers, 12 drives, etc) and want to know how to configure it. eg I'd default to 10+4 but *only* because I've been using that in my head as a placeholder for the last 2 years | 23:30 |
| notmyname | I really don't know what's best or how to tell people what to test or what to test myself | 23:30 |
| *** kei_yama has joined #openstack-swift | 23:32 | |
| kota_ | tsg: ok, FYI, cshwede also have a simulator script with GUI for replica model build by javascrit. | 23:33 |
| tsg | notmyname: valid point - let me see if I can come up with a draft for a guideline that we can turn into a doc section | 23:33 |
| notmyname | tsg: cool, that would be great! | 23:33 |
| tsg | kota_: that's cool .. link? | 23:34 |
| *** bill_az has quit IRC | 23:34 | |
| kota_ | tsg: wait...I'm now looking for it in my watching list... | 23:34 |
| kota_ | tsg: found, here https://github.com/enovance/swift-durability-calculator | 23:35 |
| tsg | kota_: ok .. let's see if we can extend this | 23:36 |
| *** bill_az has joined #openstack-swift | 23:38 | |
| kota_ | tsg: yup, I guess cshwede's one seems good to understand intiatively. For now, I'm planning to extend it with my calculartion model (but not started yet) | 23:38 |
| *** esker has quit IRC | 23:38 | |
| *** Vinsh has joined #openstack-swift | 23:48 | |
| *** Vinsh has quit IRC | 23:56 | |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!