| notmyname | >>> x = {'x':'eval(open(\'/etc/passwd\',\'rb\').read())'} | 00:04 |
|---|---|---|
| notmyname | >>> x | 00:04 |
| notmyname | {'x': "eval(open('/etc/passwd','rb').read())"} | 00:04 |
| notmyname | >>> import pickle | 00:04 |
| notmyname | >>> pickle.dumps(x) | 00:04 |
| notmyname | '(dp0\nS\'x\'\np1\nS"eval(open(\'/etc/passwd\',\'rb\').read())"\np2\ns.' | 00:04 |
| notmyname | >>> pickle.loads(pickle.dumps(x)) | 00:04 |
| notmyname | {'x': "eval(open('/etc/passwd','rb').read())"} | 00:04 |
| notmyname | torgomatic: like that? | 00:04 |
| *** lcurtis has quit IRC | 00:04 | |
| torgomatic | notmyname: I have no idea what just happened here | 00:05 |
| briancline | togormatic: yeah, I thought so. I knew I remembered seeing the xattr discussion a while back but couldn't find it, and I had to satisfy a bit of a red herring question about it for someone at work | 00:05 |
| torgomatic | you mean that strings remain strings even after a run through pickle? | 00:05 |
| notmyname | ya. ie the eval didn't happen | 00:05 |
| notmyname | but I'm not sure the right way to exploit the pickle thing | 00:06 |
| torgomatic | yup | 00:06 |
| torgomatic | really, if you have RW access to Swift data on disk, you can craft a pickle to gain control of the Swift daemons, which you can then use to... modify data on disk? | 00:06 |
| *** Nadeem has quit IRC | 00:06 | |
| torgomatic | it's sort of pointless | 00:06 |
| swifterdarrell | torgomatic: NEFARIOUS! | 00:07 |
| torgomatic | the memcache thing would let you get from "access to internal network" to "control of Swift daemons", and that's an actual escalation so we fixed it | 00:07 |
| briancline | yeah. if you have disk access, why even bother with pickling an exploit anyway | 00:08 |
| torgomatic | exactly. | 00:08 |
| briancline | thanks for enduring the question. just checking off my "yes I asked your question in the community" box | 00:10 |
| *** lihkin has quit IRC | 00:11 | |
| torgomatic | np :) | 00:11 |
| *** dmorita has joined #openstack-swift | 00:34 | |
| *** ho has joined #openstack-swift | 00:41 | |
| notmyname | for those who have logged on since my morning, note that there is a registration link in the topic for the hackathon | 00:43 |
| notmyname | space is limited, so sign up early if you're coming | 00:44 |
| *** abhirc has joined #openstack-swift | 00:51 | |
| *** ahonda has joined #openstack-swift | 01:03 | |
| mattoliverau | notmyname: so looks like I'm super effiecent... I don't remember signing up earlier, but looks like I did! So now I have 2 tickets, way for the eventbrite app to notice. | 01:05 |
| notmyname | mattoliverau: heh. I think you can change it | 01:09 |
| notmyname | mattoliverau: otherwise you're saying that we'll have to up the limit just to account for you registering twice! | 01:09 |
| mattoliverau | notmyname: No I just have to take up twice as much space and be twice as productive :p | 01:10 |
| *** abhirc has quit IRC | 01:10 | |
| notmyname | but if you are twice as productive and only twice as big, then your effective "productivity for unit matt" is actually constant. you need to be 4 times as effective if you take up 2 slots! | 01:12 |
| mattoliverau | notmyname: Lol, damn, better figure out how to cancel one then :p | 01:13 |
| mattoliverau | Turns out you can only cancel tickets from the web interface NOT the app (for those playing along at home) | 01:15 |
| mattoliverau | Lol eventbrite says your new office is 100± miles away... I guess that is a correct statement, but its also 1+ mile away :p | 01:17 |
| notmyname | and also 1000+ miles away | 01:18 |
| *** km has joined #openstack-swift | 01:18 | |
| *** abhirc has joined #openstack-swift | 01:19 | |
| *** lpabon has quit IRC | 01:20 | |
| *** zigo_ has joined #openstack-swift | 01:21 | |
| *** abhirc has quit IRC | 01:21 | |
| *** abhirc has joined #openstack-swift | 01:21 | |
| *** zigo has quit IRC | 01:21 | |
| *** kei_yama has joined #openstack-swift | 01:21 | |
| *** zigo has joined #openstack-swift | 01:27 | |
| *** zigo_ has quit IRC | 01:27 | |
| charz | mahatic | 01:31 |
| *** abhirc has quit IRC | 01:42 | |
| *** zigo has quit IRC | 01:46 | |
| notmyname | mattoliverau: FWIW I only see one ticket with your name on it | 01:47 |
| notmyname | so maybe you deleted it? | 01:47 |
| *** zigo has joined #openstack-swift | 01:47 | |
| *** addnull has joined #openstack-swift | 01:48 | |
| mattoliverau | notmyname: yeah, I cancelled it, turns out you can cancel it from the web UI but not the APP | 01:50 |
| *** abhirc has joined #openstack-swift | 01:52 | |
| *** bill_az has quit IRC | 01:54 | |
| charz | notmyname: mahatic I'm going to turn jenkins to silent mode until swift community cluster online. Sorry for inconvenience. | 01:56 |
| mahatic | charz, okay. thanks for letting me know | 01:59 |
| *** zigo has quit IRC | 02:00 | |
| *** zigo has joined #openstack-swift | 02:00 | |
| *** tellesnobrega has quit IRC | 02:06 | |
| ho | good morning! | 02:15 |
| charz | ho: good morning | 02:17 |
| openstackgerrit | Yuan Zhou proposed openstack/swift: Update contianer sync to use internal client https://review.openstack.org/143791 | 02:30 |
| *** lcurtis has joined #openstack-swift | 02:44 | |
| *** addnull has quit IRC | 03:08 | |
| *** erlon has quit IRC | 03:30 | |
| *** echevemaster has joined #openstack-swift | 03:39 | |
| *** mahatic has quit IRC | 03:48 | |
| *** fandi has joined #openstack-swift | 03:51 | |
| *** erlon has joined #openstack-swift | 03:53 | |
| *** erlon has quit IRC | 03:58 | |
| *** erlon has joined #openstack-swift | 04:17 | |
| *** abhirc has quit IRC | 04:44 | |
| *** lcurtis has quit IRC | 04:48 | |
| *** zaitcev has quit IRC | 04:59 | |
| openstackgerrit | Parantap Roy proposed openstack/python-swiftclient: SwiftClient object beginning with / or ./ https://review.openstack.org/148791 | 05:11 |
| *** SkyRocknRoll has joined #openstack-swift | 05:11 | |
| *** echevemaster has quit IRC | 05:18 | |
| *** nshaikh has joined #openstack-swift | 05:48 | |
| *** erlon has quit IRC | 05:50 | |
| *** addnull has joined #openstack-swift | 05:59 | |
| *** tellesnobrega_ has quit IRC | 06:22 | |
| *** silor has joined #openstack-swift | 06:29 | |
| *** fandi has quit IRC | 06:42 | |
| *** oomichi has quit IRC | 06:59 | |
| *** dmsimard_away is now known as dmsimard | 07:13 | |
| *** dmsimard is now known as dmsimard_away | 07:31 | |
| openstackgerrit | Parantap Roy proposed openstack/python-swiftclient: SwiftClient object beginning with / or ./ https://review.openstack.org/148791 | 07:40 |
| *** aix has joined #openstack-swift | 07:45 | |
| *** chlong has quit IRC | 07:46 | |
| *** oomichi has joined #openstack-swift | 07:52 | |
| *** ppai has joined #openstack-swift | 08:24 | |
| *** geaaru has joined #openstack-swift | 08:27 | |
| *** oomichi has quit IRC | 08:28 | |
| *** addnull has quit IRC | 08:30 | |
| *** addnull has joined #openstack-swift | 08:31 | |
| *** addnull has joined #openstack-swift | 08:31 | |
| openstackgerrit | Daisuke Morita proposed openstack/swift: Output logs of policy index https://review.openstack.org/136995 | 08:32 |
| *** kei_yama has quit IRC | 08:41 | |
| *** km has quit IRC | 08:43 | |
| *** ppai has quit IRC | 08:52 | |
| *** ppai has joined #openstack-swift | 09:06 | |
| *** acoles_away is now known as acoles | 09:08 | |
| *** jistr has joined #openstack-swift | 09:09 | |
| *** addnull has quit IRC | 09:14 | |
| *** addnull has joined #openstack-swift | 09:15 | |
| *** addnull has quit IRC | 09:18 | |
| *** fandi has joined #openstack-swift | 09:20 | |
| openstackgerrit | Daisuke Morita proposed openstack/swift: Show each policy's information on audited results in recon https://review.openstack.org/138697 | 09:25 |
| *** SkyRocknRoll has quit IRC | 09:40 | |
| *** jordanP has joined #openstack-swift | 09:46 | |
| *** jistr has quit IRC | 09:50 | |
| *** ppai has quit IRC | 09:52 | |
| *** jistr has joined #openstack-swift | 09:56 | |
| *** ppai has joined #openstack-swift | 10:05 | |
| *** addnull has joined #openstack-swift | 10:08 | |
| *** foexle has joined #openstack-swift | 10:08 | |
| *** jordanP has quit IRC | 10:09 | |
| *** addnull has quit IRC | 10:11 | |
| *** mkerrin has joined #openstack-swift | 10:12 | |
| *** ho has quit IRC | 10:12 | |
| *** tellesnobrega_ has joined #openstack-swift | 10:15 | |
| openstackgerrit | Daisuke Morita proposed openstack/swift: Show each policy's information on audited results in recon https://review.openstack.org/138697 | 10:17 |
| *** addnull has joined #openstack-swift | 10:18 | |
| *** silor has quit IRC | 10:18 | |
| *** jordanP has joined #openstack-swift | 10:23 | |
| openstackgerrit | Donagh McCabe proposed openstack/swift: Add multiple reseller prefixes and composite tokens https://review.openstack.org/137086 | 10:25 |
| *** addnull has quit IRC | 10:30 | |
| *** addnull has joined #openstack-swift | 10:32 | |
| *** nellysmitt has joined #openstack-swift | 10:32 | |
| *** jordanP has quit IRC | 10:37 | |
| *** aix has quit IRC | 10:45 | |
| *** jordanP has joined #openstack-swift | 10:50 | |
| *** rledisez has joined #openstack-swift | 10:54 | |
| *** tellesnobrega_ has quit IRC | 11:13 | |
| *** aix has joined #openstack-swift | 11:16 | |
| *** tellesnobrega has joined #openstack-swift | 11:47 | |
| *** ppai has quit IRC | 11:59 | |
| *** chlong has joined #openstack-swift | 12:02 | |
| *** saltsa has quit IRC | 12:02 | |
| *** saltsa has joined #openstack-swift | 12:03 | |
| *** ppai has joined #openstack-swift | 12:13 | |
| *** ppai has quit IRC | 12:22 | |
| *** ppai has joined #openstack-swift | 12:34 | |
| *** nellysmitt has quit IRC | 13:01 | |
| *** silor has joined #openstack-swift | 13:06 | |
| *** fandi has quit IRC | 13:08 | |
| *** erlon has joined #openstack-swift | 13:22 | |
| *** abhirc has joined #openstack-swift | 13:28 | |
| *** ppai has quit IRC | 13:39 | |
| *** abhirc has quit IRC | 13:45 | |
| *** bill_az has joined #openstack-swift | 13:48 | |
| *** ppai has joined #openstack-swift | 13:48 | |
| *** abhirc has joined #openstack-swift | 13:49 | |
| *** abhirc has quit IRC | 13:52 | |
| *** mikehn has quit IRC | 13:54 | |
| *** tellesnobrega has quit IRC | 13:54 | |
| *** mikehn has joined #openstack-swift | 13:57 | |
| *** tellesnobrega has joined #openstack-swift | 14:06 | |
| *** ppai has quit IRC | 14:07 | |
| *** mahatic has joined #openstack-swift | 14:08 | |
| *** dmsimard_away is now known as dmsimard | 14:31 | |
| *** lcurtis has joined #openstack-swift | 14:35 | |
| *** lcurtis has quit IRC | 14:35 | |
| *** dmsimard is now known as dmsimard_away | 14:36 | |
| *** lihkin has joined #openstack-swift | 14:37 | |
| *** nshaikh has quit IRC | 14:47 | |
| *** addnull has quit IRC | 14:50 | |
| *** nellysmitt has joined #openstack-swift | 15:02 | |
| *** zaitcev has joined #openstack-swift | 15:04 | |
| *** ChanServ sets mode: +v zaitcev | 15:04 | |
| *** nellysmitt has quit IRC | 15:06 | |
| *** jasondotstar has joined #openstack-swift | 15:08 | |
| openstackgerrit | Takashi Kajinami proposed openstack/swift: Add process name checking into swift-init https://review.openstack.org/116203 | 15:08 |
| *** jasondotstar has quit IRC | 15:08 | |
| *** jasondotstar has joined #openstack-swift | 15:09 | |
| *** dmsimard_away is now known as dmsimard | 15:11 | |
| *** lpabon has joined #openstack-swift | 15:13 | |
| *** tdasilva has joined #openstack-swift | 15:13 | |
| *** abhirc has joined #openstack-swift | 15:22 | |
| *** jasondotstar has quit IRC | 15:38 | |
| *** lihkin has quit IRC | 15:41 | |
| *** lihkin has joined #openstack-swift | 15:43 | |
| openstackgerrit | Mahati proposed openstack/swift: Implement OPTIONS verb for storage nodes. https://review.openstack.org/140103 | 15:52 |
| *** fandi has joined #openstack-swift | 15:57 | |
| *** lcurtis has joined #openstack-swift | 15:57 | |
| *** jasondotstar has joined #openstack-swift | 16:01 | |
| *** lihkin has quit IRC | 16:08 | |
| *** lihkin has joined #openstack-swift | 16:09 | |
| *** briancurtin has joined #openstack-swift | 16:14 | |
| *** abhirc has quit IRC | 16:25 | |
| *** foexle has quit IRC | 16:26 | |
| *** jasondotstar has quit IRC | 16:27 | |
| *** abhirc has joined #openstack-swift | 16:33 | |
| notmyname | good morning | 16:34 |
| openstackgerrit | paul luse proposed openstack/swift: Merge master to feature/ec https://review.openstack.org/148983 | 16:39 |
| peluse | clayg, torgomatic : FYI the merge above resolves a conflict between a few of your patches... please take a quick look (not urgent though) | 16:42 |
| peluse | morning! | 16:42 |
| *** nellysmitt has joined #openstack-swift | 16:48 | |
| *** lcurtis has quit IRC | 16:51 | |
| *** lcurtis has joined #openstack-swift | 16:57 | |
| notmyname | https://wiki.openstack.org/wiki/Swift/PriorityReviews is updated with a section for the next release. just a few things there (I'd love to see them land soon) | 16:57 |
| notmyname | anyone have anything for the meeting today? that ^ is all I had | 16:58 |
| notmyname | oh, and the hackathon (link in topic) | 16:58 |
| *** tongli has joined #openstack-swift | 17:03 | |
| notmyname | mahatic: around? | 17:06 |
| *** atan8 has joined #openstack-swift | 17:09 | |
| *** atan8 has quit IRC | 17:10 | |
| *** atan8 has joined #openstack-swift | 17:11 | |
| *** rledisez has quit IRC | 17:15 | |
| *** mjfork has joined #openstack-swift | 17:17 | |
| mjfork | I have a node in my cluster that all I reduced all of the drive weights by 20% relative to the other nodes but they continue to show the same or higher capacity used than the higher weighted counterparts. Since dropping we have completed multiple replication cycles. Any suggestions on where to start looking? | 17:19 |
| mahatic | notmyname, sorry! | 17:20 |
| mahatic | notmyname, around now | 17:20 |
| mahatic | notmyname, I was cooking and lost track of time :P | 17:20 |
| notmyname | mjfork: no worries. what were you cooking? | 17:20 |
| notmyname | mjfork: what version of swift are you running? | 17:21 |
| mahatic | notmyname, just some bread and eggs | 17:21 |
| notmyname | mahatic: no worries. what were you cooking? <-- now with the right nick :-) | 17:21 |
| mjfork | 1.12.0 | 17:21 |
| mahatic | notmyname, lol yes | 17:22 |
| mahatic | notmyname, did you happen to take a look at the patch? I updated it for the rest of the two containers | 17:24 |
| notmyname | mahatic: since my review yesterday? | 17:27 |
| notmyname | looking now | 17:27 |
| mahatic | yeah | 17:27 |
| *** echevemaster has joined #openstack-swift | 17:27 | |
| mahatic | i mean rest of the two servers | 17:30 |
| notmyname | yup. looks good | 17:30 |
| notmyname | good catch by ppai | 17:30 |
| mahatic | yeah! | 17:30 |
| notmyname | mahatic: ok, since this one looks ready to go, then the next step (while you're waiting for this to be reviewed and land) is to work on the small patch to add the server type header | 17:31 |
| notmyname | mahatic: how do you think that should look? | 17:31 |
| mahatic | I also have a question on that, in the test - test_call_incorrect_replication_method, if I replace the obj_methods with 'REPLICATE' | 17:31 |
| mahatic | notmyname, obj/server.py returns 200, but account and container return 400 - bad request. It's a value error and I think it is because of the input that is being passed to REPLICATE method | 17:33 |
| notmyname | hmm | 17:33 |
| mahatic | am i correct? | 17:33 |
| notmyname | let me check | 17:33 |
| mjfork | notmyname: just realized i didn't tag you in my answer - we are running 1.12.0 (I know...back level) | 17:34 |
| notmyname | mjfork: I saw. I didn't have an immediate answer. sorry :-) | 17:34 |
| notmyname | mjfork: how "balanced" are your failure domains? | 17:35 |
| notmyname | mjfork: ie looking at total available raw capacity in each region/zone/server, how even is it? | 17:35 |
| mjfork | notmyname: very unbalanced, we are gradually adding 2x the number of nodes (about 20% in at this point) | 17:36 |
| notmyname | mjfork: ah | 17:36 |
| notmyname | mahatic: ok, just to talk through it... | 17:41 |
| notmyname | mahatic: if you set up the server to be a replication server, then give a REPLICATE verb. the object server gives 200 and the account/container give 400 | 17:42 |
| notmyname | right? | 17:42 |
| notmyname | mjfork: that's ... not unexpected. as in, I'm not surprised to hear that answer, and there's been some work recently to fix some of the data placement when there are very uneven failure domains | 17:43 |
| mahatic | notmyname, yeah, in the current scenario itself - replication:true, so if I go head replace the verbs with REPLICATE, that's what I get (the result you just mentioned) | 17:43 |
| notmyname | mjfork: note I'm not saying that it's good. just that you aren't seeing a "special snowflake" situation. probably | 17:43 |
| mahatic | replication_server:true | 17:44 |
| *** jordanP has quit IRC | 17:44 | |
| *** jasondotstar has joined #openstack-swift | 17:44 | |
| mjfork | notmyname: thanks. what i expected. | 17:44 |
| notmyname | mjfork: you're on the east coast of the US, right? | 17:45 |
| mjfork | notmyname: yes | 17:45 |
| notmyname | mjfork: ok. mahatic is in India, so her timezone is way off from us. let me finish working with her, and then I can help you out (more details or whatever you need) | 17:46 |
| mjfork | notmyname: ok, works for me. thanks. | 17:46 |
| notmyname | mahatic: I'm looking at the differences in implementation | 17:47 |
| mahatic | notmyname, sure | 17:47 |
| *** gvernik has joined #openstack-swift | 17:53 | |
| openstackgerrit | Donagh McCabe proposed openstack/swift-specs: Minor updates to composite token spec https://review.openstack.org/138771 | 17:53 |
| mahatic | notmyname, mjfork thanks for working out the timings! | 17:53 |
| notmyname | mahatic: what you're seeing makes sense, I was just reminding myself of the "why" | 17:53 |
| *** lihkin has quit IRC | 17:54 | |
| mahatic | notmyname, hmm, json.load(req.environ['wsgi.input']) where wsgi.input = stringIO() doesn't go right I guess | 17:55 |
| notmyname | mahatic: so the object server expects /drive/part/suffix/policy and the account+container servers expects /drive/partition/hash | 17:55 |
| *** lihkin has joined #openstack-swift | 17:56 | |
| mahatic | yup | 17:56 |
| notmyname | and since the test has "/sdb1/p/a/c" for all of the test paths, the account+container servers fail with 400 because of the bad path | 17:56 |
| *** jistr has quit IRC | 17:56 | |
| notmyname | mahatic: so, technically, the requests you're building in the tests for account+container are wrong. however, what you're testing is that the other verbs respond with 405. so it still tests the right thing | 17:57 |
| notmyname | mahatic: does that make sense? do you agree? | 17:57 |
| mahatic | notmyname, yes, it makes sense. Why the existing is working. But not the sdb1/p/a/c part. If it's a bad path, why is not effecting the object server? | 17:57 |
| notmyname | mahatic: because the "is this a valid verb" check happens first, before the "is this a valid path" check | 17:58 |
| mahatic | existing test* | 17:58 |
| notmyname | mahatic: the path parsing is done _in_ the verb implementation | 17:58 |
| mahatic | notmyname, correct, so it should fail for object server too, correct? | 17:59 |
| mahatic | am I missing something? | 18:00 |
| notmyname | mahatic: no, because the fake path you have happens to match the pattern the object server's REPLICATE verb expects, and the implementation of the object's REPLCIATE verb ends up not failing if the given thing referenced ont he path doesn't exist | 18:01 |
| mahatic | notmyname, okay | 18:02 |
| *** tongli has quit IRC | 18:02 | |
| mahatic | notmyname, so the existing test does make the correct check, should I write one more test each for account + container that contains the correct path? | 18:03 |
| notmyname | ah. the account+container servers are expecting json in the request body | 18:05 |
| mahatic | yes and hence I thought it was a ValueError | 18:05 |
| notmyname | mahatic: well, maybe. but that would be a different test. right now, your test checks that other verbs respond with 405 if replication server is true | 18:06 |
| notmyname | so that's a good test | 18:06 |
| mahatic | okay | 18:06 |
| notmyname | but what you're talking about is a different check that the replicate verb does something different | 18:06 |
| notmyname | (I'm guessing there's already some sort of test for that) | 18:06 |
| *** acoles is now known as acoles_away | 18:07 | |
| mahatic | hmm | 18:07 |
| notmyname | hmmm indeed | 18:08 |
| notmyname | initial check says "no" | 18:08 |
| mahatic | ah | 18:08 |
| mahatic | and another check says no too | 18:11 |
| notmyname | looks like there are several tests on the object server | 18:13 |
| notmyname | so at least there's that :-/ | 18:13 |
| mahatic | yup | 18:13 |
| notmyname | mahatic: ok, so it seems that there is a gap there | 18:14 |
| notmyname | and yup. completely confirmed. unittests never execute those lines in the container server (so says the coverage report) | 18:14 |
| mahatic | ah | 18:15 |
| mahatic | I see | 18:15 |
| notmyname | mahatic: however, while important, it's orthogonal to the OPTIONS/recon stuff you're working on | 18:15 |
| notmyname | but thanks for pointing it out :-) | 18:15 |
| mahatic | so it ends up being a different patch/work altogether? | 18:16 |
| mahatic | okay! no problem :) | 18:17 |
| notmyname | ya, I think so | 18:17 |
| notmyname | mahatic: what are your next steps? do you know yet? | 18:18 |
| mahatic | notmyname, so getting back to the existing patch, server type header? | 18:18 |
| notmyname | yes | 18:18 |
| mahatic | notmyname, er I don't understand what it is about | 18:18 |
| mahatic | where should I make a change? any reference? | 18:19 |
| notmyname | mahatic: ok, at least on the OPTIONS response (but maybe on all of them--but that might be harder), there should be a header returned saying what kind of server it is | 18:19 |
| notmyname | mahatic: so there are 2 parts: 1) set the value and 2) return the value | 18:19 |
| *** geaaru has quit IRC | 18:20 | |
| mahatic | notmyname, okay | 18:20 |
| notmyname | mahatic: setting is easy: have a default in the base class and then set it in each child class (ie a class attribute) | 18:20 |
| notmyname | mahatic: for returning it, it should be trivially simple to set the header on the OPTIONS response. you can find many many places in the code on how to set headers on a response, if you don't remember how to do it off the top of your head | 18:21 |
| notmyname | mahatic: but the question remains as to what header to set (ie the header name) | 18:21 |
| mahatic | notmyname, okay. don't we already have something like X-Server-Type? | 18:22 |
| notmyname | mahatic: so to figure that out, I'd like you to do some research and find if there is a standard header (eg defined in an RFC) that we should use | 18:22 |
| notmyname | we don't have anything yet | 18:22 |
| mahatic | okay | 18:22 |
| notmyname | so we should use something standard if there is something appropriate | 18:22 |
| notmyname | maybe "Server". I remember something about that somewhere, but I don't remember details | 18:23 |
| mahatic | okay | 18:23 |
| mahatic | will look into that | 18:23 |
| *** atan8 has quit IRC | 18:23 | |
| notmyname | mahatic: if you don't find anything standard that we should use, then you get to invent on. eg X-Server-Type or X-Swift-Server or X-Mahatis-Header or whatever | 18:23 |
| mahatic | notmyname, :D yeah, last one sounds cool, so i'll go with that ;) | 18:24 |
| notmyname | lol | 18:24 |
| mahatic | before whatever that is :D | 18:24 |
| notmyname | point is, you set the header and then, finally (next step after this), you read that header with the recon tool | 18:24 |
| *** jasondotstar has quit IRC | 18:24 | |
| notmyname | mahatic: now for extra bonus points, the header could be returned with every response from the server. but figure out just doing the OPTIONS response first | 18:25 |
| mahatic | oh okay | 18:25 |
| notmyname | mahatic: make sensse? does that give you enough to go on for now? | 18:25 |
| *** Nadeem has joined #openstack-swift | 18:25 | |
| mahatic | notmyname, sure, I believe so. I'll do a quick check to confirm I understood things, or I'll post some questions here | 18:25 |
| notmyname | mahatic: great! | 18:27 |
| mahatic | notmyname, I see that x-delete-at or X-Backend-Timestamp kind of headers being set in the class, are they all defined or created in some config file? | 18:28 |
| notmyname | where? | 18:29 |
| mahatic | in container/server.py -> gen_resp_headers method -> | 18:30 |
| mahatic | X-Backend-Timestamp, X-Backend-PUT-Timestamp etc are set | 18:31 |
| notmyname | ya. doesn't look like that's used for every response, but that's similar to how it'd be done for every response | 18:32 |
| mahatic | no, I'm not saying it's used for every response. I wanted to know where are they all initially defined | 18:32 |
| *** nshaikh has joined #openstack-swift | 18:33 | |
| mahatic | they are not standard headers if I'm not wrong? So aren't they defined/created somewhere? | 18:33 |
| notmyname | the only way headers are "defined" is that they get added to a headers dictionary somewhere. so no there isn't a formal "here's how to define a header" thing anywhere. | 18:34 |
| notmyname | but on the other hand, yes, that may be where those are used and set (and thus defined) for that server | 18:35 |
| notmyname | does that make sense? | 18:35 |
| mahatic | yeah where do I find that headers dictionary? I'll just have to navigate through the code and find out? | 18:36 |
| mahatic | notmyname, that's what I meant to find out. If there is any common place for this sort of a dictionary | 18:38 |
| notmyname | ah | 18:39 |
| notmyname | maybe. I mean, that would be nice. probably not though | 18:39 |
| notmyname | the gen_resp_headers() is probably one of the closest things | 18:39 |
| mahatic | okay, no problem, will look through | 18:39 |
| *** annegent_ has joined #openstack-swift | 18:40 | |
| mahatic | okay | 18:40 |
| notmyname | mjfork: still around? I've got about 20 minutes now | 18:41 |
| mahatic | notmyname, thank you for the inputs. Will look through and post back any questions | 18:42 |
| notmyname | mahatic: you're welcome | 18:42 |
| *** annegent_ has quit IRC | 18:52 | |
| *** acoles_away is now known as acoles | 18:54 | |
| notmyname | good drive info from backblaze. good info when looking at selecting drives for swift too https://www.backblaze.com/blog/best-hard-drive/ | 18:54 |
| ahale | some shocking seagate numbers ! | 18:55 |
| *** cutforth has joined #openstack-swift | 18:56 | |
| notmyname | ya. looks like just the 3TB drives | 18:56 |
| notmyname | don't buy those ;-) | 18:56 |
| notmyname | interesting summary at the bottom: 4TB drives are good and based on price about equal | 18:57 |
| notmyname | swift meeting time | 19:00 |
| *** lihkin has quit IRC | 19:02 | |
| *** aix has quit IRC | 19:04 | |
| mattoliverau | Well that was short and sweet. I'm going back to bed then :) cya all in an hour or two. | 19:15 |
| notmyname | :-) | 19:16 |
| clayg | notmyname: if mjfork is on 1.12 he's in the exact situation as cschwede that caused us to let weights fight dispersion when failure domains at a tier is less than replicas - changing the weight on old code had zero effect on partition placement if you have <= failure domains at a tier than replicas in that tier | 19:17 |
| *** acoles is now known as acoles_away | 19:17 | |
| notmyname | right | 19:17 |
| notmyname | and thanks for the detail | 19:17 |
| * notmyname steps out for an early lunch | 19:22 | |
| *** mahatic has quit IRC | 19:28 | |
| *** silor has quit IRC | 19:37 | |
| *** annegent_ has joined #openstack-swift | 19:39 | |
| *** annegent_ has quit IRC | 19:44 | |
| *** fifieldt__ has joined #openstack-swift | 19:53 | |
| *** jrichli has joined #openstack-swift | 19:55 | |
| *** fifieldt_ has quit IRC | 19:56 | |
| *** pberis has quit IRC | 19:57 | |
| *** erlon has quit IRC | 20:00 | |
| *** jasondotstar has joined #openstack-swift | 20:04 | |
| *** jasondotstar has quit IRC | 20:09 | |
| *** gvernik has quit IRC | 20:18 | |
| *** nshaikh has left #openstack-swift | 20:24 | |
| *** erlon has joined #openstack-swift | 20:28 | |
| notmyname | new office space means a whole new set of lunch options | 20:29 |
| *** fandi has quit IRC | 20:29 | |
| occupant | I'm about to move into office space that's in a mall, so there'll be gigantic food court directly underneath me. | 20:30 |
| notmyname | sbarro every day? ;-) | 20:30 |
| occupant | it's a westfield mall, so the food court is pretty decent. | 20:31 |
| notmyname | oh, nice | 20:31 |
| notmyname | occupant: how's the sound isolation? I'd expect that to be pretty noisy | 20:32 |
| occupant | it's several floors above the actual mall, so just like any normal building. my employer is providing the noise by ripping out all the cubes and replacing them with a bullshit open plan. | 20:34 |
| notmyname | ugh | 20:34 |
| occupant | because it's real fucking "collaborative" when everyone needs to have noise cancelling headphones strapped to their head just so they can hear themselves think | 20:34 |
| occupant | I'm pretty sure it just started out as a bullshit ex post facto justification to cram a bunch of people into a small space on cheap ikea desks | 20:35 |
| occupant | but people accidentally started believing it for real | 20:35 |
| notmyname | ya, I think that's pretty common. cram a lot of people in and realize it's relatively cheap and then justify it | 20:36 |
| *** nellysmitt has quit IRC | 20:37 | |
| *** nellysmitt has joined #openstack-swift | 20:38 | |
| openstackgerrit | Ian Cordasco proposed openstack/python-swiftclient: Release connection after consuming the content https://review.openstack.org/149043 | 20:39 |
| *** annegent_ has joined #openstack-swift | 20:40 | |
| *** sigmavirus24 has joined #openstack-swift | 20:40 | |
| sigmavirus24 | Hey everyone, I just submitted https://review.openstack.org/#/c/149043/ but I was wondering if anyone could provide some guidance on where the best place to add a test for that would be. Thanks! | 20:41 |
| *** mjfork has quit IRC | 20:44 | |
| *** annegent_ has quit IRC | 20:45 | |
| *** bpap has joined #openstack-swift | 20:47 | |
| *** fandi has joined #openstack-swift | 20:48 | |
| *** fandi has quit IRC | 20:53 | |
| *** fandi has joined #openstack-swift | 20:53 | |
| *** omame has quit IRC | 20:58 | |
| *** fandi has quit IRC | 20:58 | |
| *** omame has joined #openstack-swift | 20:58 | |
| dmsimard | Seeing a lot of ChunkWriteTimeout by swift-proxy from image uploads through glance on a virtual CI setup. I'm not expecting everything to be super performant but is there anything I should be especially looking at ? | 20:58 |
| *** fandi has joined #openstack-swift | 20:58 | |
| *** tdasilva has quit IRC | 21:07 | |
| zaitcev | The fundamental problem, as I understand, is that Swift API is HTTP and clients time out eventually. They cannot know if servers are loaded or not. So, say they close after a minute, which is actually a lot for libraries that support browsers. | 21:08 |
| zaitcev | In that minute, the proxy must time out duff nodes and retry. | 21:09 |
| zaitcev | Therefore, nodes have to have a latency that's less than a minute. In that there are DB ops with sqlite. | 21:09 |
| zaitcev | Long story short, you start seeing the above when service times on disks start climbing. | 21:10 |
| zaitcev | If utilization breaks 80% and elevator starts working, you see queue going into 10 sec, and it's game over | 21:11 |
| zaitcev | So Swift nodes cannot live on overloaded CI machines. They just can't. | 21:11 |
| zaitcev | notmyname et.al. should correct my understanding here, but generally it's what I see from attempts to place Swift on busy VM hosts. | 21:12 |
| *** tdasilva has joined #openstack-swift | 21:20 | |
| *** jeblair has joined #openstack-swift | 21:22 | |
| *** fandi has quit IRC | 21:26 | |
| *** tdasilva has quit IRC | 21:28 | |
| openstackgerrit | Ian Cordasco proposed openstack/python-swiftclient: Release connection after consuming the content https://review.openstack.org/149043 | 21:39 |
| *** annegent_ has joined #openstack-swift | 21:39 | |
| *** annegent_ has quit IRC | 21:44 | |
| *** david-lyle has joined #openstack-swift | 21:46 | |
| *** mjfork has joined #openstack-swift | 21:47 | |
| *** fandi has joined #openstack-swift | 21:47 | |
| *** sigmavirus24 has left #openstack-swift | 21:55 | |
| mjfork | clayg: sorry for delay ... we are on 1.12 and we have 20+ zones and only 3 replicas. Does that mean the issue you mentioned would not apply? | 21:56 |
| *** chlong has quit IRC | 22:05 | |
| *** nellysmitt has quit IRC | 22:08 | |
| *** nellysmitt has joined #openstack-swift | 22:09 | |
| *** nellysmitt has quit IRC | 22:09 | |
| *** wasmum has left #openstack-swift | 22:10 | |
| *** david-lyle has quit IRC | 22:11 | |
| *** david-lyle has joined #openstack-swift | 22:11 | |
| clayg | hrmmm.... yeah with >3 zones in the same region 1.12 should acknowledge the lower weight in one of the zones | 22:14 |
| clayg | mjfork: ^ | 22:15 |
| clayg | mjfork: maybe i misunderstood your description of the issue - you reduced the weight of the devices in one of the zones and the part count went down but you're not seeing replication move those parts off and free up disk space? | 22:15 |
| *** david-lyle has quit IRC | 22:15 | |
| dmsimard | zaitcev: Just saw your reply - thanks | 22:20 |
| briancline | clayg: yeah, they were reduced in a zone by about 20% and it didn't seem to free up much of any space on devices within that zone | 22:22 |
| clayg | briancline: did parts move in balance, and just replication was slow/not-working? Or is rebalance not moving parts off the de-weighted devices? | 22:25 |
| clayg | briancline: mjfork: those two problems are kind of different - swift-ring-builder something.builder should tell us how many parts are assigned to the devices - it should be ~20% less than the other devices (a factor of weight) | 22:25 |
| briancline | clayg: checking on the part counts for those, one moment | 22:30 |
| clayg | hey hey! gil's back! https://review.openstack.org/#/c/143791/ | 22:30 |
| *** tellesnobrega_ has joined #openstack-swift | 22:32 | |
| notmyname | clayg: yup. he was chatting in here earlier | 22:32 |
| *** annegent_ has joined #openstack-swift | 22:39 | |
| *** erlon has quit IRC | 22:40 | |
| clayg | what the fuck is up with testr!? I can't even tell what test is failing!? | 22:41 |
| *** tellesnobrega_ has quit IRC | 22:41 | |
| clayg | why is there so much output? | 22:41 |
| portante | clayg: not to be of any help here, but testr likes to run tests and tell you all about them! | 22:42 |
| *** annegent_ has quit IRC | 22:44 | |
| clayg | apparently there's like a follow on command that knows how to make sense of the output maybe? | 22:45 |
| clayg | '\n'.join('I hate python-swiftclient's test suite.' for i in range(100)) | 22:49 |
| *** tellesnobrega_ has joined #openstack-swift | 22:49 | |
| *** mjfork has quit IRC | 22:51 | |
| clayg | time tox -e py27 tests.unit.test_service.TestServiceUtils.test_process_options_defaults ... Ran 1 tests in 0.044s (-0.002s) ... real1m3.785s | 22:53 |
| clayg | ^ how can we be expected to live like this!? | 22:53 |
| clayg | a'ignt nobody got time for that!? | 22:54 |
| notmyname | clayg: http://www.quickmeme.com/img/ef/ef5c72d07e23f4e1423e708c874cb54114f9f89beae1ce019c35a6066604d43a.jpg | 22:55 |
| clayg | this better be a trick to make tox run faster | 22:55 |
| clayg | ... i'll give you half credit | 22:55 |
| notmyname | :-) | 22:55 |
| *** david-lyle has joined #openstack-swift | 22:58 | |
| *** tellesnobrega_ has quit IRC | 23:00 | |
| clayg | ahhhhh..... real 0m4.492s | 23:00 |
| clayg | ah..... even better! real 0m0.779s | 23:01 |
| clayg | python -m testtools.run <test> was the answer - "thanks" https://wiki.openstack.org/wiki/Testr#FAQ | 23:01 |
| peluse | clayg, ahh, handy.. I rarely just one but when I do I've been editing tox.ini -- sheesh that's much easier. | 23:03 |
| * peluse takes note: always check the FAQs | 23:03 | |
| clayg | peluse: that was more like my first google result for 'god i hate you testr!' or somethig similar | 23:04 |
| peluse | clayg, I say that kinda shit to siri all the time but it never seems to get me results | 23:05 |
| *** david-lyle has quit IRC | 23:09 | |
| clayg | gd, i'm sure there is some combination of environment variables that will cause this test to pass... | 23:12 |
| clayg | ahhh unset ST_AUTH | 23:12 |
| clayg | that doesn't make any sense - we already patch that dictionary | 23:14 |
| *** jrichli has quit IRC | 23:20 | |
| *** tdasilva has joined #openstack-swift | 23:23 | |
| *** km has joined #openstack-swift | 23:25 | |
| *** chlong has joined #openstack-swift | 23:33 | |
| *** IRTermite has quit IRC | 23:37 | |
| *** annegent_ has joined #openstack-swift | 23:39 | |
| *** echevemaster has quit IRC | 23:43 | |
| *** tdasilva has quit IRC | 23:44 | |
| *** echevemaster has joined #openstack-swift | 23:44 | |
| *** annegent_ has quit IRC | 23:44 | |
| clayg | bah, i don't know an easy way to patch the environ before you import the module | 23:50 |
| *** oomichi_ has joined #openstack-swift | 23:50 | |
| *** dmsimard is now known as dmsimard_away | 23:57 | |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!
