*** zhill has quit IRC | 00:00 | |
openstackgerrit | Samuel Merritt proposed openstack/swift: EC: fix PUT with If-None-Match: * https://review.openstack.org/167456 | 00:02 |
---|---|---|
ho | good morning guys! | 00:04 |
mattoliverau | ho: morning | 00:09 |
ho | mattoliverau: morning! | 00:11 |
*** vinsh has joined #openstack-swift | 00:23 | |
openstackgerrit | Denis Cavalcante proposed openstack/swift: Add storage policy support for sorting method https://review.openstack.org/160877 | 00:26 |
*** kota_ has joined #openstack-swift | 00:35 | |
*** gyee has quit IRC | 00:48 | |
openstackgerrit | Merged openstack/swift: Fix common misspellings https://review.openstack.org/166843 | 00:53 |
*** shri has quit IRC | 00:54 | |
InAnimaTe | looks like one day | 00:57 |
notmyname | peluse: to be pendantic, functional tests against an *EC* policy only fully pass once a couple of outstanding patches land ;-) | 00:58 |
mattoliverau | InAnimaTe: cool, thanks for searching for it. Sorry, don't hack on python-swiftclient much.. Should remedy that someday :) | 00:58 |
mattoliverau | notmyname: so you want to go land more patches? | 00:59 |
InAnimaTe | mattoliverau: yeah it appears its just the same as tempauth since they are so simiar | 00:59 |
InAnimaTe | however, im not seeing any way to force expire a token, which forces a client to re-auth and pull a new x-storage-url, which is what i need to do since ours are presently an ip address | 01:00 |
InAnimaTe | which isn't cool | 01:00 |
notmyname | mattoliverau: that's what I'm doing tonight :-) | 01:00 |
mattoliverau | InAnimaTe: for tempauth, clear it out of the memcache and the client will have to re-auth | 01:02 |
InAnimaTe | ohh and derp of course i find it after the fact: https://github.com/gholt/swauth/blob/master/swauth/middleware.py#L128 | 01:05 |
InAnimaTe | ohhh, crap thats a good idea | 01:05 |
InAnimaTe | i dont really know how to work with memcache but is there a way i can clear only that token for specific user x? or do i just need to clear the entire cache? | 01:06 |
mattoliverau | InAnimaTe: I think you can clear individual objects.. but you'd have to google, you can like telnet in and run commands or something | 01:06 |
InAnimaTe | so | 01:08 |
InAnimaTe | swauth-cleanup-tokens | 01:08 |
InAnimaTe | ^that exists | 01:08 |
InAnimaTe | however, just running it tracebacks | 01:08 |
InAnimaTe | lol | 01:08 |
notmyname | master->ec merge is being rechecked | 01:18 |
notmyname | (again) | 01:18 |
*** tsg_ has quit IRC | 01:18 | |
mattoliverau | Yay, this time for sure :p | 01:26 |
notmyname | success! | 01:27 |
notmyname | (with the requirements job) | 01:28 |
peluse | awesome! | 01:28 |
peluse | the rest is a cake walk :) | 01:28 |
notmyname | heh | 01:28 |
openstackgerrit | Denis Cavalcante proposed openstack/swift: Add storage policy support for sorting method https://review.openstack.org/160877 | 01:32 |
yuan | peluse, mattoliverau, is everything OK of the feature/ec branch on your side? I saw a few unit tests failures if I set the default policy to a EC one | 01:44 |
*** panbalag has quit IRC | 01:45 | |
*** jkugel has joined #openstack-swift | 01:46 | |
*** kei_yama has joined #openstack-swift | 01:52 | |
*** km has joined #openstack-swift | 01:52 | |
peluse | yuan, there's a big merge that's going to land here pretty soon | 01:52 |
peluse | like approved in the next 10 min then as long as it takes to make it through | 01:53 |
*** peluse has left #openstack-swift | 01:53 | |
*** peluse has joined #openstack-swift | 01:54 | |
*** ChanServ sets mode: +v peluse | 01:54 | |
peluse | and then after that there's a ton of changes coming... | 01:54 |
*** haomaiwang has joined #openstack-swift | 01:58 | |
peluse | BTW earlier I mentioned probe, unit and func all passing on feature/ec after the merge - that's with a repl pol as def. with Ec as default there's still work to do on func but unit should all work. I'd wait til after the next big chain lands before digging into anything not directly related to something you're working on though | 02:02 |
peluse | but again, should be 0 issues with unit test after the merge lands | 02:02 |
peluse | git branch | 02:02 |
* peluse grumbles | 02:02 | |
openstackgerrit | Yuan Zhou proposed openstack/swift: Update contianer sync to use internal client https://review.openstack.org/143791 | 02:05 |
*** doxavore has joined #openstack-swift | 02:11 | |
*** agentle has joined #openstack-swift | 02:13 | |
*** dencaval has quit IRC | 02:17 | |
*** agentle has quit IRC | 02:48 | |
*** agentle has joined #openstack-swift | 02:53 | |
*** ppai has joined #openstack-swift | 02:55 | |
*** jkugel has quit IRC | 02:57 | |
*** agentle has quit IRC | 02:57 | |
*** doxavore has quit IRC | 02:59 | |
openstackgerrit | Merged openstack/swift: Merge master to feature/ec https://review.openstack.org/167429 | 03:12 |
mattoliverau | yay | 03:27 |
kota_ | great | 03:27 |
*** kei_yama_ has joined #openstack-swift | 03:28 | |
*** kei_yama has quit IRC | 03:28 | |
notmyname | hello, again | 03:31 |
notmyname | yuan: still here? | 03:33 |
notmyname | yuan: please rebase https://review.openstack.org/#/c/162890/ when you get a chance | 03:37 |
*** mikehn has quit IRC | 03:38 | |
*** mikehn has joined #openstack-swift | 03:38 | |
openstackgerrit | Samuel Merritt proposed openstack/swift: EC: fix PUT with If-None-Match: * https://review.openstack.org/167456 | 03:42 |
notmyname | torgomatic: rebase? | 03:43 |
torgomatic | notmyname: yup | 03:43 |
yuan | notmyname, sure | 03:44 |
notmyname | yuan: thanks :-) | 03:44 |
peluse | I'll start on "the chain" in the morning unless some elves take care of it tonight.... over and out | 03:56 |
notmyname | peluse: sleep well. tomorrow's another big day ;-) | 03:58 |
notmyname | interesting. S3 just added cross-region replication https://aws.amazon.com/blogs/aws/new-cross-region-replication-for-amazon-s3/ | 03:59 |
kota_ | notmyname: oh, really? | 03:59 |
kota_ | interesting... | 04:00 |
*** km_ has joined #openstack-swift | 04:03 | |
*** km has quit IRC | 04:04 | |
openstackgerrit | Yuan Zhou proposed openstack/swift: Fix copy from different type policy https://review.openstack.org/162890 | 04:07 |
*** tsg has joined #openstack-swift | 04:07 | |
notmyname | torgomatic: if you're still around tonight, can you validate (with your +2) that yuan's rebase is still ok in your estimation? https://review.openstack.org/#/c/162890/ | 04:17 |
*** ppai has quit IRC | 04:27 | |
*** kota_ has quit IRC | 04:29 | |
*** km has joined #openstack-swift | 04:30 | |
*** km_ has quit IRC | 04:32 | |
*** ppai has joined #openstack-swift | 04:41 | |
*** sandywalsh_ has quit IRC | 04:52 | |
*** reed has quit IRC | 05:00 | |
*** sandywalsh has joined #openstack-swift | 05:04 | |
notmyname | well that's cool | 05:06 |
notmyname | with https://review.openstack.org/#/c/162890/ applied on top of https://review.openstack.org/#/c/167456/, every functional test against a cluster with a default ec policy passes! | 05:07 |
zaitcev | nice | 05:08 |
notmyname | yuan: if the patches land in that order, then there may need to be a conflict resolution. I did it locally for testing, and it's really simple | 05:11 |
*** zaitcev has quit IRC | 05:25 | |
*** tsg has quit IRC | 05:29 | |
*** welldannit has quit IRC | 05:34 | |
yuan | notmyname, OK I 'll watch out for that | 05:36 |
*** kota_ has joined #openstack-swift | 05:43 | |
*** SkyRocknRoll has joined #openstack-swift | 05:44 | |
*** zhill has joined #openstack-swift | 05:49 | |
*** zhill has quit IRC | 05:49 | |
*** jamielennox is now known as jamielennox|away | 05:49 | |
openstackgerrit | Hisashi Osanai proposed openstack/swift: Clarify the description of backward compatibility in Keystoneauth's docstring https://review.openstack.org/166681 | 05:57 |
yuan | kota_, is there a link for the shss backend? | 06:13 |
kota_ | yuan: src? | 06:13 |
kota_ | yuan: or binary? | 06:13 |
yuan | both are OK | 06:14 |
kota_ | yuan: neither of those, sorry. | 06:14 |
*** bkopilov has joined #openstack-swift | 06:15 | |
kota_ | yuan: currently, it's our proprietary product. | 06:15 |
kota_ | yuan: so there is no online link to shss engine. | 06:16 |
yuan | oh I see. I was doing something tests with liberasurecode here so want to have a try | 06:16 |
kota_ | yuan: I see, sorry for the inconvenience... | 06:17 |
yuan | kota_, never mind | 06:17 |
kota_ | yuan: However, liberasurecode tests works well in my env with/without shss | 06:18 |
kota_ | yuan: it will show something like "backend library is not avairable" when shss didn't be installed. | 06:19 |
kota_ | yuan: If you want to make some changes for pyeclib/liberasurecode, please push it to the bitbucket repository. If so, I could test with shss at my local env. | 06:21 |
*** nshaikh has joined #openstack-swift | 06:21 | |
yuan | yeah, also working on my env. I was tyring to make some small benchmark tools for liberasurecode so just want to have all the lib ready | 06:22 |
kota_ | yuan: sounds great :) Let's push it to upstream ;) | 06:23 |
yuan | kota_, yeah I will push out later this week. | 06:29 |
openstackgerrit | Hisashi Osanai proposed openstack/swift: Enable Object Replicator's failure count in recon https://review.openstack.org/138342 | 06:32 |
*** silor has joined #openstack-swift | 07:00 | |
openstackgerrit | Merged openstack/swift: Fix copy from different type policy https://review.openstack.org/162890 | 07:10 |
*** Novtopro_ has joined #openstack-swift | 07:26 | |
*** Novtopro_ has quit IRC | 07:30 | |
*** Novtopro_ has joined #openstack-swift | 07:31 | |
*** ppai has quit IRC | 07:35 | |
*** Novtopro_ has quit IRC | 07:40 | |
*** ppai has joined #openstack-swift | 07:49 | |
*** Bsony has joined #openstack-swift | 07:50 | |
*** chlong has quit IRC | 07:52 | |
openstackgerrit | Yuan Zhou proposed openstack/swift: EC: fix PUT with If-None-Match: * https://review.openstack.org/167456 | 07:55 |
*** jordanP has joined #openstack-swift | 07:59 | |
*** joeljwright has joined #openstack-swift | 08:01 | |
*** mmcardle has joined #openstack-swift | 08:18 | |
ho | yuan: I could not see any difference on above patch. could you teach me it? | 08:19 |
ho | yuan: #167456 | 08:20 |
yuan | ho, yes, it's only some minor rebasing work | 08:22 |
yuan | otherwise there's a merge conflict | 08:22 |
ho | yuan: i see. thanks :) | 08:25 |
*** jistr has joined #openstack-swift | 08:38 | |
*** km has quit IRC | 08:45 | |
*** geaaru has joined #openstack-swift | 08:47 | |
*** kei_yama_ has quit IRC | 08:47 | |
*** Bsony_ has joined #openstack-swift | 08:51 | |
*** Bsony has quit IRC | 08:54 | |
*** ho has quit IRC | 08:59 | |
*** srsakhamuri has quit IRC | 09:15 | |
cschwede | acoles_away: is it ok for you if i fix the two minor things mentioned by Kota on patch 165208 (logging/unused variable)? Don’t want to delay things, and I would like to put a +2 afterwards on it | 09:17 |
patchbot | cschwede: https://review.openstack.org/#/c/165208/ | 09:17 |
*** ppai has quit IRC | 09:20 | |
*** Bsony_ has quit IRC | 09:22 | |
*** acoles_away is now known as acoles | 09:34 | |
*** ppai has joined #openstack-swift | 09:34 | |
acoles | cschwede: hi! re logging, i confess i did the same as the finalize_put method to be consistent. I did wonder why that called logging.exception and not the logger - can you see any reason? | 09:39 |
acoles | cschwede: see line 933 in the patchset diskfile.py | 09:40 |
cschwede | acoles: i think that could be changed as well, i’m happy to submit a patch to use self._diskmanager.logger there too. just noticed that Sam introduced self.logger in ECAppIter as well, thus we could make it consistent to use self.logger / self.manager.logger everywhere | 09:44 |
cschwede | thus said, i see no reason not to use self.manager.logger | 09:44 |
*** kota_ has quit IRC | 09:44 | |
acoles | cschwede: ah, i see clayg made the change at line 1783 do its consistent to make the one other change at line 1799 | 09:45 |
acoles | cschwede: (i'm just catching up with last nights rebases/changes!) | 09:45 |
acoles | cschwede: so if you are happy to push over those changes then please do or i can do it this morning. PLease don't change finalize_put logging though - we are trying to leave the legacy DiskFile classes as untouched as possible in the ec patches | 09:47 |
cschwede | acoles: ok, i only change _finalize_durable (the one kota commented on as well) | 09:49 |
acoles | cschwede: there's one other nit still not fixed too - test_write_durable_fsync --> test_commit_fsync line 2139 test_diskfile.py | 09:49 |
cschwede | shall i rename it? | 09:49 |
acoles | cschwede: yes please! or you want me to do the changes? | 09:50 |
cschwede | acoles: already done :) | 09:50 |
openstackgerrit | Christian Schwede proposed openstack/swift: Add commit method to interface of DiskFile https://review.openstack.org/165208 | 09:50 |
acoles | cschwede: so 3 changes in total - the logger call, remove expected=1 and the test method name | 09:50 |
cschwede | https://review.openstack.org/#/c/165208/7..8/swift/obj/diskfile.py | 09:51 |
cschwede | https://review.openstack.org/#/c/165208/7..8/test/unit/obj/test_diskfile.py | 09:51 |
cschwede | that’s all i changed | 09:51 |
acoles | cschwede: fast! i guess you are 1 hour ahead of me :) | 09:51 |
*** joeljwright1 has joined #openstack-swift | 09:52 | |
acoles | cschwede: thanks! | 09:52 |
*** joeljwright has quit IRC | 09:53 | |
cschwede | acoles: you’re welcome! | 09:53 |
*** navid__ has joined #openstack-swift | 09:54 | |
*** navid__ has quit IRC | 09:54 | |
*** navid__ has joined #openstack-swift | 09:55 | |
*** nshaikh has quit IRC | 09:55 | |
*** navid__ has quit IRC | 09:55 | |
*** nshaikh has joined #openstack-swift | 09:55 | |
openstackgerrit | Christian Schwede proposed openstack/swift: EC: better failure handling for GETs https://review.openstack.org/167406 | 09:58 |
openstackgerrit | Christian Schwede proposed openstack/swift: EC: support multiple ranges for GET requests https://review.openstack.org/166576 | 09:58 |
acoles | notmyname: joeljwright1 : just reading your discussion about a swiftclient release in my scrollback. i agree we should cut a release with what we have, there's good stuff still to come but there is good stuff there already so lets get it out the door and press on with reviewing joeljwright1 other patches post-EC. | 09:59 |
joeljwright1 | acoles: notmyname cut 2.4.0 last night | 09:59 |
joeljwright1 | acoles: quite pleased to have some of those patches in a released version :) | 10:04 |
joeljwright1 | acoles: you might also want to take a look at this patch post-EC https://review.openstack.org/#/c/161043/ | 10:10 |
joeljwright1 | looks really useful | 10:10 |
mattoliverau | acoles, cschwede: just took a quick look at patch 165208 and wondered if the write_durable in the test function names should now also called commit as well? I've only +1'ed for now based on that.. happy to +2 it once gerrit has come back and told me its good (it's late and I'm lazy) :P | 10:17 |
patchbot | mattoliverau: https://review.openstack.org/#/c/165208/ | 10:17 |
cschwede | mattoliverau: good catch - I could quickly submit another patchset if you want (and acoles is ok with it too) | 10:19 |
mattoliverau | sure, if you do, I | 10:19 |
mattoliverau | *i'll run tox and if all good with give it my +2 | 10:19 |
mattoliverau | cschwede: ^ | 10:20 |
openstackgerrit | Christian Schwede proposed openstack/swift: Add commit method to interface of DiskFile https://review.openstack.org/165208 | 10:23 |
cschwede | mattoliverau: acoles: done | 10:23 |
cschwede | i quickly grepped trough the code, couldn’t find any other traces of write_durable | 10:24 |
mattoliverau | cschwede: nice, looks good. Running a tox now and as soon as jenkins comes back and has passed I'll +2/+A it :) | 10:26 |
*** ppai has quit IRC | 10:42 | |
*** Bsony has joined #openstack-swift | 10:48 | |
acoles | cschwede: mattoliverau oh were there other test method names we missed :/ thanks for sorting those out! loving this globally distributed effort :) | 10:52 |
*** ppai has joined #openstack-swift | 10:55 | |
mattoliverau | acoles: it is pretty cool :) | 10:58 |
acoles | joeljwright1: hi! thx for the update (i only scanned scrollback quickly!) i added myself on 161043 so i keep track of it | 11:00 |
*** admin6 has joined #openstack-swift | 11:01 | |
openstackgerrit | Christian Schwede proposed openstack/swift: EC: support multiple ranges for GET requests https://review.openstack.org/166576 | 11:14 |
openstackgerrit | Christian Schwede proposed openstack/swift: EC: better failure handling for GETs https://review.openstack.org/167406 | 11:16 |
mattoliverau | Cools, +A'ed. On that note, I'm going to bed. Night all | 11:17 |
acoles | mattoliverau: thanks! g'night | 11:18 |
*** dmorita has quit IRC | 11:18 | |
*** panbalag has joined #openstack-swift | 11:31 | |
*** a1|away is now known as AbyssOne | 11:33 | |
*** ppai has quit IRC | 11:33 | |
*** AbyssOne is now known as a1|away | 11:34 | |
*** a1|away is now known as AbyssOne | 11:35 | |
openstackgerrit | Merged openstack/swift: Handle ENOSPC in mkstemp() https://review.openstack.org/161713 | 11:35 |
*** ppai has joined #openstack-swift | 11:45 | |
*** cdelatte has joined #openstack-swift | 11:49 | |
*** chlong has joined #openstack-swift | 11:50 | |
peluse | hey, I like what I'm reading. acoles - how goes the battle? | 11:55 |
acoles | peluse: hi. well we have two patches heading to landing this morning on the diskfile chain | 11:56 |
peluse | nice! | 11:56 |
peluse | maybe I should go back to bed! | 11:57 |
acoles | peluse: i'm reviewing the next one in the chain right now (clayg's hash suffixy stuff) | 11:57 |
acoles | peluse: where else should i be paying some attention? any gaps in the effort? | 11:58 |
peluse | the chain should be job #1 right now so sounds like guys have been rocking it | 11:58 |
admin6 | Hi guys, I’d like to know what is the status of EC development and if it is planned to be included as stable version in kilo ? | 11:58 |
peluse | admin6, it will be in kil as a beta | 11:59 |
acoles | admin6: we are working really hard towards that goal right now | 11:59 |
peluse | and depending on feedback the plan will be to release production when its ready :) | 12:00 |
admin6 | I know that you’re working reaaly hard :-) | 12:00 |
peluse | so later this year. we need lots of performance feedback, likely some tuning - few bug fixes along the way. that kinda stuff. Its a pretty big change | 12:00 |
acoles | peluse: so i stay focussed on the hash sufix/ssync/reconstructor chain? | 12:00 |
admin6 | ‘when its ready’ is a good timeframe | 12:00 |
peluse | acoles, perfect! | 12:00 |
acoles | peluse: i think we made some progress with getting policy configurable func tests to work yesterday too, so i'm pretty hopeful we can guide that in by end of week | 12:01 |
peluse | I'm going to stay on the reconstructor but holla if you need me | 12:02 |
peluse | man, the good news keeps coming :) | 12:02 |
cschwede | i’m reviewing patch 159637 atm, will finish the review/tests later. ping me if someone needs another review for EC today | 12:02 |
patchbot | cschwede: https://review.openstack.org/#/c/159637/ | 12:02 |
acoles | cschwede: great, i'm looking at 159637 too so if we're both happy we may get that approved today too. | 12:03 |
acoles | peluse: i'll go look for some bad news to balance things out ;) | 12:03 |
peluse | cschwede, thanks for the efforts here lately - really have made a difference | 12:03 |
cschwede | peluse: you’re welcome! | 12:04 |
*** openstackgerrit has quit IRC | 12:06 | |
*** openstackgerrit has joined #openstack-swift | 12:07 | |
*** haigang has joined #openstack-swift | 12:11 | |
*** haigang has quit IRC | 12:13 | |
*** haigang has joined #openstack-swift | 12:14 | |
*** haigang has quit IRC | 12:14 | |
*** happyeveryday has joined #openstack-swift | 12:14 | |
tdasilva | good morning | 12:17 |
tdasilva | wow, looks like you guys got a lot done overnight :-) | 12:17 |
tdasilva | sweet, patch 167429 landed. I will rebase 164950 | 12:18 |
patchbot | tdasilva: https://review.openstack.org/#/c/167429/ | 12:18 |
tdasilva | acoles: do you have a sec to talk about func. tests? | 12:19 |
tdasilva | I guess Janie is not around yet? | 12:19 |
tdasilva | acoles: I started to create a patch for the functional tests on master with the change to select we talked about, but then I realized that for the CrossPolicy testing, we might need to allow the selection of the default policy, in the case there are only two policies defined. | 12:21 |
tdasilva | acoles: so it first tries to pick the one defined in the env. variable, but if that is already being used, then it may select another random one (which could be the default one) | 12:22 |
*** admin6_ has joined #openstack-swift | 12:29 | |
*** admin6 has quit IRC | 12:30 | |
*** admin6_ is now known as admin6 | 12:30 | |
acoles | tdasilva: hi. thanks for following up on that. i think its fine to select the default for the second policy. | 12:45 |
acoles | tdasilva: maybe i confused you in discussion yesterday | 12:46 |
tdasilva | acoles: nah, it's just that when you actually start working on the code, some things become more clear :-) | 12:47 |
tdasilva | acoles: in fact, when I took a step back, I started wondering why not just have people change the default policy when running func tests? | 12:47 |
tdasilva | we are just trying to override the default policy, right? | 12:48 |
acoles | tdasilva: at risk of chinese whispers distorting someone else's intent :) ... i think the goal was to make it easier to switch between policy type under test, not having to hack config and restart cluster. Plus, you may not have access to cluster to change default policy when running func tests | 12:50 |
*** Bsony has quit IRC | 12:50 | |
acoles | tdasilva: so one day we might have devstack configured with multiple policies, then run a jenkins job for func test using each policy in turn | 12:50 |
acoles | *one day* ! | 12:51 |
tdasilva | acoles: ok, that's what I figured...it's just that coming from me just running these tests on my SAIO vm, I started wondering why would I ever use this | 12:51 |
acoles | s/devstack/devstack swift/ | 12:51 |
tdasilva | acoles: but yeah, that makes sense | 12:51 |
tdasilva | acoles: thanks for the clarification :-) | 12:55 |
*** ppai has quit IRC | 12:55 | |
*** Bsony has joined #openstack-swift | 12:56 | |
acoles | tdasilva: no problem | 12:56 |
openstackgerrit | Merged openstack/swift: Cleanup ECDiskFile classes https://review.openstack.org/167097 | 12:59 |
openstackgerrit | Merged openstack/swift: Add commit method to interface of DiskFile https://review.openstack.org/165208 | 13:00 |
openstackgerrit | Prashanth Pai proposed openstack/swift: Make object creation more atomic in Linux https://review.openstack.org/162243 | 13:01 |
peluse | yes! | 13:02 |
*** jkugel has joined #openstack-swift | 13:04 | |
*** jkugel1 has joined #openstack-swift | 13:05 | |
*** jkugel has quit IRC | 13:09 | |
*** Bsony has quit IRC | 13:16 | |
*** foexle has joined #openstack-swift | 13:17 | |
*** foexle has quit IRC | 13:17 | |
*** foexle has joined #openstack-swift | 13:17 | |
*** Bsony has joined #openstack-swift | 13:21 | |
*** jrichli has joined #openstack-swift | 13:23 | |
*** Bsony has quit IRC | 13:29 | |
*** doxavore has joined #openstack-swift | 13:30 | |
doxavore | should one expect the auditor to consistently do a lot _writing_? like a 1:7 read:write ratio, all the time? | 13:35 |
*** SkyRocknRoll has quit IRC | 13:36 | |
doxavore | (as measured in MB/s and i/o utilization) | 13:36 |
*** openstackgerrit has quit IRC | 13:36 | |
*** openstackgerrit has joined #openstack-swift | 13:36 | |
*** Bsony has joined #openstack-swift | 13:38 | |
*** annegentle has joined #openstack-swift | 13:42 | |
openstackgerrit | Merged openstack/swift: EC: fix PUT with If-None-Match: * https://review.openstack.org/167456 | 13:48 |
*** wasmum has quit IRC | 13:48 | |
jrichli | tdasilva acoles: how is the functest change going? I saw the exchange from earlier today | 13:54 |
tdasilva | jrichli: hi, I was about to send a patch to master, but I'd be happy to just send you a patch in case you want to submit it yourself | 13:55 |
jrichli | oh, so you are making a new review on master? | 13:56 |
jrichli | tdasilva: you made more changes, though, on top of what I had done. so I would think it should have your name. | 13:58 |
tdasilva | jrichli, acoles: yeah, I was just following on the idea from acoles that this is generic enough that could be on master | 14:00 |
openstackgerrit | Thiago da Silva proposed openstack/swift: Select policy when running functional test https://review.openstack.org/167595 | 14:00 |
acoles | cschwede: i am looking at https://review.openstack.org/#/c/159637/25/test/unit/obj/test_diskfile.py and think that some of the tests have been duplicated, e.g. test_hash_cleanup* | 14:00 |
tdasilva | jrichli: ^ check it out and please feel free to take over as you see fit | 14:00 |
jrichli | tdasilva: sure, I will take a look. thanks! | 14:01 |
jrichli | tdasilva: thanks for the reference in the commit message :-) | 14:02 |
tdasilva | jrichli: it's really nothing more than your patch, plus the change to select, so you deserve all the credit! | 14:02 |
acoles | cschwede: when i reviewed version 20 the originals in TestDiskFileModuleMethods had been removed but now they are back in version 25, so I am thinking there's been a mistake | 14:02 |
acoles | cschwede: so similar tests appear in TestDiskFileModuleMethods and TestSuffixHashes | 14:03 |
acoles | cschwede: i'll have a go at cleaning it up then wait to see what clayg has to say later | 14:03 |
cschwede | acoles: hmm, yes, i think you’re right. you mean the tests in line 258 ff in https://review.openstack.org/#/c/159637/20..25/test/unit/obj/test_diskfile.py , right? | 14:05 |
acoles | cschwede: yes | 14:06 |
acoles | cschwede: ok i'll try to fix it without breaking it ;) | 14:07 |
*** uschreiber_ has joined #openstack-swift | 14:08 | |
*** lpabon has joined #openstack-swift | 14:10 | |
cschwede | acoles: let me know if i can help out | 14:11 |
*** mahatic has joined #openstack-swift | 14:13 | |
*** nshaikh has left #openstack-swift | 14:14 | |
*** Bsony has quit IRC | 14:22 | |
*** vinsh has quit IRC | 14:23 | |
*** wasmum has joined #openstack-swift | 14:34 | |
*** vinsh has joined #openstack-swift | 14:44 | |
*** reed has joined #openstack-swift | 14:47 | |
openstackgerrit | Thiago da Silva proposed openstack/swift: Validate the PUT method extraction for EC https://review.openstack.org/164950 | 14:51 |
tdasilva | clayg: ^^^proxy refactoring rebased | 14:52 |
*** tsg_ has joined #openstack-swift | 14:54 | |
*** zaitcev has joined #openstack-swift | 14:57 | |
*** ChanServ sets mode: +v zaitcev | 14:57 | |
petertr7 | Is there anyone working on or interested in an opensource version of swiftstack? | 15:06 |
openstackgerrit | Alistair Coles proposed openstack/swift: Multiple Fragment Archive support for suffix hashes https://review.openstack.org/159637 | 15:06 |
acoles | cschwede: ^^ | 15:06 |
tdasilva | acoles: got a python question: whenever I try to move the import statement to the top in this file https://review.openstack.org/#/c/167595/1/test/functional/swift_test_client.py I get this error: AttributeError: 'module' object has no attribute 'functional' | 15:06 |
tdasilva | acoles: any ideas? | 15:06 |
tdasilva | jrichli: did you run into this issue too? | 15:07 |
jrichli | tdasilva: yes | 15:07 |
*** mahatic has quit IRC | 15:08 | |
jrichli | tdasilva: that is why I put the import where it is. It is because there is a cyclical dependency, I think | 15:08 |
*** openstackgerrit has quit IRC | 15:08 | |
*** uschreiber_ has quit IRC | 15:09 | |
*** openstackgerrit has joined #openstack-swift | 15:09 | |
tdasilva | jrichli: mm..i see | 15:10 |
acoles | tdasilva: looking... | 15:11 |
tdasilva | acoles, jrichli: I need to step out for a bit, but will be back later...for now I'm sending a new patchset with acoles comments applied minus this import issue | 15:13 |
jrichli | tdasilva: ok, thx | 15:14 |
straycat | is the ring building/rebalancing process deterministic? | 15:16 |
*** happyeveryday has quit IRC | 15:16 | |
straycat | so, can i run it on 3 nodes separately with the same options to obtain the same configs rather than running on one node and distributing to the rest? | 15:17 |
*** happyeveryday has joined #openstack-swift | 15:20 | |
openstackgerrit | Thiago da Silva proposed openstack/swift: Select policy when running functional test https://review.openstack.org/167595 | 15:20 |
cschwede | straycat: no, there are some randomly choosen values in it for the distribution. it’s not deterministic, and thus you have to distribute the same set of ringfiles to all storage/proxy nodes | 15:21 |
acoles | tdasilva: jrichli : yes there is a cycle there, best ask one of the python gurus what the best practice is | 15:22 |
acoles | clayg: i need to you to doublecheck what i have done on https://review.openstack.org/159637 - it looks like there was some old code crept back in to v25, and i gambled that was not what you intended, but maybe you did?? | 15:27 |
straycat | cschwede, thanks | 15:38 |
*** mahatic has joined #openstack-swift | 15:39 | |
clayg | acoles: looking | 15:48 |
clayg | cschwede: you can pass in a seed tho | 15:50 |
clayg | acoles: for some reason the 25->26 change has a bunch of noise? | 15:50 |
clayg | acoles: i guess the master merge went down | 15:50 |
*** annegentle has quit IRC | 15:51 | |
cschwede | clayg: right, but then you need to rebalance the ring on all nodes instead doing it only once, and i think the whole process might be error-prone | 15:51 |
clayg | cschwede: yeah well... i'm not saying it's a good idea - just that if you tried it and it didn't work you could always file a bug - because it's *supposed* to work that way | 15:53 |
clayg | acoles: looks good to me - let's merge everything | 15:54 |
cschwede | clayg: agreed, very true | 15:55 |
*** Bsony has joined #openstack-swift | 15:56 | |
clayg | acoles: oh, you guys already did :\ | 15:56 |
*** annegentle has joined #openstack-swift | 15:57 | |
clayg | wow - great work guys - I love the "here I'll just push up a new change set - ftfy +2" style reviews! | 15:59 |
*** Bsony has quit IRC | 16:02 | |
clayg | tdasilva: thanks for the help with patch 164950 - your last message said you're stepping out so I'm going to fix up some jenkins errors | 16:02 |
patchbot | clayg: https://review.openstack.org/#/c/164950/ | 16:02 |
notmyname | good morning. just read scrollback | 16:02 |
notmyname | straycat: you can give the ring builder a seed value for the PRNG so that it is deterministic across machines (/cc cschwede) | 16:03 |
acoles | clayg: 25->26 noise is just a few cleanup in the patches that landed earlier | 16:04 |
straycat | okay, what about rebalancing? | 16:04 |
clayg | notmyname: everyone is pickin' on poor cschwede this morning - he's all like trying to be the voice of reason "WHY would you do that to yourself?!" and we keep saying "yeah but you *could* do it" | 16:04 |
straycat | i realise what i'm trying to do isn't exactly the standard route | 16:04 |
clayg | straycat: I think we normally call that "crazytown" | 16:04 |
clayg | but that's ok - we *like* crazy | 16:05 |
straycat | :') | 16:05 |
notmyname | lol. the reason it was implemented was because a deployer wanted to build rings locally on all hosts. they submitted a patch. (then they stopped using swift altogether *grumble*) | 16:05 |
clayg | notmyname: i like having the see for tests - that about it | 16:05 |
notmyname | ya, it helps for tests too :-) | 16:05 |
*** bkopilov has quit IRC | 16:06 | |
notmyname | oh, heh. clayg I just saw that you had already said the same thing to cschwede | 16:06 |
clayg | notmyname: maybe it's a sign that if straycat tries to do this he'll eventually decide swift's to hard? | 16:06 |
straycat | sorry, so ring creation is deterministic, is rebalancing? | 16:07 |
clayg | there is some pseudo randomness that can be forced to be deterministic with a seed | 16:07 |
notmyname | straycat: there isn't actually a difference there. it's all the same method under the covers. ie the seed value applies | 16:07 |
straycat | okay | 16:08 |
straycat | i basically need to automate the deployment of identical nodes, and i'd like to do this without having to make one node distribute the ring files to the others | 16:08 |
straycat | so that's why i'm asking | 16:08 |
clayg | straycat: so you have to co-ordinate the seed when rebalancing - but other than that it'll work out | 16:08 |
clayg | straycat: and after the initial deployment how are you going to handle capacity management? | 16:08 |
clayg | where to you co-ordinate how the identical nodes get a list of all nodes - can the guy who distributes that distribute the ring's? | 16:09 |
straycat | clayg, i don't follow your last question | 16:10 |
straycat | in this particular case i'm not expecting the deployment to change over time | 16:10 |
straycat | the set of devices will stay the same | 16:10 |
clayg | heh | 16:10 |
cschwede | straycat: sounds like you have something like puppet,chef,ansible whatever in place? you could use that tool to deploy the ring instead of the ring settings? | 16:11 |
clayg | that would be *awesome* - can I buy some of these never failing devices from you? | 16:11 |
notmyname | lol | 16:11 |
clayg | notmyname: no think about it - the drives never change - that would be a game changer - you don't even need a balancing algo - you just write out replica2part2dev and be done with it | 16:12 |
straycat | cschwede, exactly, i've got a tool that will that will run the ring builder and copy the result onto the node's filesystem, then deploy that filesystem to hardware | 16:13 |
cschwede | notmyname: interesting to get to know the history of these little details (seed for ring builder). i only assumed it was for testing purposes only | 16:13 |
clayg | notmyname: why can't I think of that guys name... McSomething? | 16:14 |
clayg | er. Mac? worked with the dickhead that's at pivitol labs now | 16:14 |
clayg | oh McKenty - so what was the other guy? | 16:14 |
clayg | Chris? | 16:14 |
*** jkugel1 has quit IRC | 16:15 | |
clayg | McGowan - how did I never notice before they were *both* Mc! | 16:17 |
notmyname | clayg: be nice | 16:17 |
clayg | :D | 16:18 |
*** chlong has quit IRC | 16:22 | |
cschwede | straycat: so, assuming none of your disks will ever change it could work if you use the same seed value. that said i still think there are better/safer ways. simplest example would be a file on a puppetmaster that gets distributed to the nodes (same mechanism would apply to other deyploment tools) | 16:22 |
cschwede | straycat: have a look at https://github.com/stackforge/puppet-swift/tree/master/manifests if you by chance use puppet | 16:22 |
straycat | cschwede, am using baserock for this ( http://wiki.baserock.org/ ) | 16:25 |
cschwede | straycat: ah, i have an assumption about your idea. makes sense to use a deterministic ring building process then | 16:29 |
straycat | cool, thanks | 16:30 |
cschwede | straycat: assuming your nodes might be located on different networks, make sure the storage network uses a vpn (if there are public networks between the nodes) | 16:30 |
*** jordanP has quit IRC | 16:31 | |
*** jordanP has joined #openstack-swift | 16:31 | |
clayg | i don't get why the disk image based approach means that deterministic ring building makes more sense than just distributing the ring.gz on image instead of the .builder? | 16:32 |
notmyname | clayg: I'm cleaning up what's starred in gerrit | 16:32 |
notmyname | I think I've got the root of the various patch chains there | 16:32 |
notmyname | clayg: but what about https://review.openstack.org/#/c/164950/ ? | 16:32 |
*** chlong has joined #openstack-swift | 16:33 | |
clayg | notmyname: that's the one i'm working on right now! | 16:33 |
clayg | notmyname: wait, what's the question? | 16:33 |
straycat | clayg, it seems simpler, to distribute i have to wait for the nodes to boot and bring up ssh or something | 16:33 |
notmyname | clayg: shall I star it? is it needed for beta? | 16:33 |
notmyname | ie is it one that needs to be reviewed asap? | 16:34 |
acoles | clayg: re patch 159637 i can't see what test replaces test_hash_suffix_one_file i.e. a test that passes reclaim_age to hash_suffix and tests .ts is reclaimed? | 16:34 |
patchbot | acoles: https://review.openstack.org/#/c/159637/ | 16:34 |
straycat | cschwede, why, does seeding make the setup vulnerable? | 16:34 |
clayg | notmyname: oh - yes please star - should be reviewable shortly | 16:34 |
clayg | acoles: oh no! | 16:35 |
acoles | clayg: the new tests all call df_mgr.get_hashes which doesn't pass reclaim_age ?? | 16:35 |
acoles | clayg: warning - my eyes are really tired today!) | 16:35 |
clayg | acoles: i tried to be careful :\ | 16:35 |
clayg | acoles: rather than passing in reclaim age I probably would have written a test that just created an old tombstone and made sure the hash got all reset and the suffix was reaped | 16:36 |
clayg | acoles: but maybe I didn't write that test yet | 16:36 |
clayg | acoles: I *should* - ding me on the review | 16:36 |
openstackgerrit | Clay Gerrard proposed openstack/swift: Validate the PUT method extraction for EC https://review.openstack.org/164950 | 16:37 |
acoles | clayg: yes thats the kind of test i've been hunting for - something that creates a .ts at (now-ONE_WEEK) | 16:37 |
notmyname | jrichli: if work is going on master for https://review.openstack.org/#/c/166097/ please abandon this one. I want to keep feature/ec open patches focused | 16:38 |
clayg | acoles: cool i'm on it - can you try and find some more? | 16:38 |
clayg | stupid coffee cup must have a hole in it | 16:39 |
*** chlong has quit IRC | 16:39 | |
jrichli | notmyname: done | 16:40 |
notmyname | jrichli: thanks | 16:40 |
jrichli | np! | 16:41 |
acoles | clayg: i'm trying my best :) we do test that HCL reclaims, but I just removed the reclaim_age arg from hash_suffix and all test_diskfile.py tests passed so think there is a gap there. | 16:42 |
*** jkugel has joined #openstack-swift | 16:42 | |
clayg | acoles: we test that the module function reclaims - or the manager function | 16:42 |
clayg | acoles: what about the quarantined suffix dir that's in the red just right above the one_file test? where did it go? | 16:43 |
clayg | i would have gone in order when I was working on them | 16:44 |
notmyname | ok, gerrit cleaned up. I love all of your patches equally, but right now I love https://review.openstack.org/#/c/159637/ more equally. that looks liek the one that needs the most time | 16:44 |
clayg | lol | 16:44 |
acoles | clayg: mgr, in new tests its line 2900 test_hash_cleanup_listdir_purge_old_ts | 16:44 |
notmyname | ok, omw to the office | 16:44 |
cschwede | actually patch 159637 is quite nice already | 16:45 |
patchbot | cschwede: https://review.openstack.org/#/c/159637/ | 16:45 |
acoles | clayg: we have test_hash_suffix_hash_cleanup_listdir_enotdir_quarantined that i think covers what was the quarantine test in red | 16:45 |
acoles | clayg: basically i'm going through the old one by one and finding what in the new replaces it | 16:46 |
acoles | clayg: so far i'm at 4 out of a gazillion :D | 16:46 |
clayg | acoles: yeah that's awesome - so you think test_hash_cleanup_listdir_purge_old_ts is a good replacement for test_hash_suffix_one_file | 16:46 |
straycat | cschwede, okay i assume yes, sorry for the dumb question i just wanted to be sure i understood your suggestion | 16:46 |
cschwede | straycat: seeding doesn’t make it vulnerable; i am more worried that some nodes apply a wrong update, for example when a newer swift version is rolled out, but one node uses an old one for example. thus i feel more comfortable when the ring is built in one place and the .ring.gz are shipped | 16:47 |
straycat | cschwede, right, we can also assume the swift version won't change | 16:47 |
acoles | clayg:no because it didn't fail when i did this http://paste.openstack.org/show/196674/ | 16:48 |
clayg | acoles: so that's in the module level function | 16:48 |
*** gyee has joined #openstack-swift | 16:49 | |
clayg | acoles: oh - never mind ok i got it | 16:49 |
acoles | clayg: it doesn;t test the hash_suffix path | 16:49 |
acoles | clayg: k, i'll comment and continue with audit | 16:50 |
*** jistr has quit IRC | 16:58 | |
clayg | acoles: do you think running a test in a for loop to make sure it's deterministic is "doing it right" or does that mean I'm doing something way wrong? | 16:58 |
acoles | clayg: not sure what you mean? you mean repeat test over and over and check result is always the same? | 16:59 |
*** zhill has joined #openstack-swift | 17:02 | |
acoles | clayg: i like that you added some more meaning to the test names | 17:03 |
clayg | acoles: YOU FOUND A BUG! | 17:05 |
clayg | the way I had plumbed _get_hashes it always used the default reclaim age | 17:06 |
clayg | acctully this bug may have been there when i started | 17:07 |
*** admin6 has left #openstack-swift | 17:08 | |
*** annegentle has quit IRC | 17:11 | |
*** annegentle has joined #openstack-swift | 17:12 | |
acoles | clayg: plumbing to _get_hashes looks the same on master - the mgr get_hashes doesn't pass a reclaim age, looks like just the calls from replicator | 17:12 |
acoles | direct to module function | 17:12 |
clayg | figures | 17:12 |
clayg | good think EC doesn't use the replicator | 17:13 |
clayg | ? | 17:13 |
*** vinsh_ has joined #openstack-swift | 17:13 | |
clayg | acoles: wait - so do you think I shouldn't change it? | 17:13 |
acoles | clayg: change what ? | 17:13 |
acoles | the plumbing? | 17:14 |
clayg | should get hashes pass in the reconfigured reclaim age? | 17:14 |
clayg | well... i mean of course it should? | 17:14 |
clayg | what if you set your reclaim age to a month - it can't just go useing the default like a mad man! | 17:14 |
*** vinsh has quit IRC | 17:16 | |
acoles | the config sample only has reclaim_age option for replicator (and reconstructor) ?? | 17:17 |
*** vinsh has joined #openstack-swift | 17:17 | |
acoles | # The replicator also performs reclamation | 17:18 |
acoles | # reclaim_age = 604800 | 17:18 |
acoles | so is that meant to suggest that the object-server doesn't reclaim? | 17:18 |
clayg | idk, it's not like the default is "dont reclaim anything" - it's reclaim anything older than one week - it doesn't make *any* sense to me for one process to use one value and another something different | 17:18 |
acoles | clayg: agree, and if i set that option to one month i'd expect to see some .ts hanging around for 4 weeks | 17:19 |
clayg | fo' sure | 17:19 |
acoles | clayg: so what was this about deterministic testing? | 17:20 |
clayg | i'm going to fix it in DiskFileManager.get_hashes - I think it's most obvious there that it has a self.reclaim_age option and it's not using | 17:20 |
clayg | *it* | 17:20 |
*** lpabon has quit IRC | 17:20 | |
*** vinsh_ has quit IRC | 17:20 | |
clayg | acoles: do you ever run the same unittest in a for loop - for i in {1..10}; do !!; if [ $? -ne 0]; then break; fi; done | 17:21 |
acoles | clayg: yes if i thought the test was non-deterministic meaning it has random seed values or something | 17:22 |
clayg | i do this a fair amount - and rather than making me think i'm being diligent - sometimes it just makes me feel like I don't trust the tests i'm writing :D | 17:22 |
*** cdelatte has quit IRC | 17:23 | |
clayg | acoles: i've taken to doing it in any test that has any calls to "time" in it anywhere :P | 17:23 |
acoles | we have tests that choose random conditions , idk, what i don't like about that is they aren't repeatable | 17:23 |
clayg | maybe not quite that bad | 17:23 |
acoles | so if we see a test failure we do a recheck and it goes away and we move on | 17:23 |
acoles | ideally i'd have every test be repeatable/deterministic, try to cover all scenarios, and then also have randomly seeded "soak tests" that log their starting state if they fail | 17:24 |
acoles | like, in an ideal world | 17:25 |
clayg | soak test - i like it | 17:26 |
acoles | clayg: case in point - https://bugs.launchpad.net/swift/+bug/1435538 what happened there??? | 17:26 |
openstack | Launchpad bug 1435538 in OpenStack Object Storage (swift) "intermittent ratelimit unit test failure" [Low,New] | 17:26 |
clayg | acoles: I think maybe the Mixin's might help us cover more types of policies with less test maintance | 17:26 |
acoles | clayg: i'm over the mixin hate, I put +2+A on that :P | 17:27 |
acoles | but yes, i *like* the pattern | 17:27 |
acoles | i'm channeling my angst at my ide today | 17:28 |
clayg | well i don't want it to distrupt your workflow - I think the abc thing would probably work - but I don't have pycharm installed to test it - so it's all you | 17:28 |
clayg | acoles: just stick a few def assertX on the mixin and see if it doesn't shut up? | 17:28 |
clayg | acoles: anyway that bug looks like your average run of the mill gate was slow bug? that test only has 100ms varience on it - i'm surprised it doesn't fail more often than it does? | 17:30 |
clayg | acoles: you want me to push this over -> https://gist.github.com/clayg/44b341b9e131289c2603 | 17:31 |
clayg | acoles: you could also go ahead and make your comments and I'll address what you have - or just wait patiently | 17:31 |
acoles | clayg: i'm not sure about making the reclaim_age change in this patch? shouldn't that go on master? idk | 17:34 |
clayg | of course it *should* :P | 17:34 |
acoles | clayg: you slipped a _one_ into test_hash_suffix_one_reclaim_tombstone | 17:34 |
acoles | clayg: :P | 17:34 |
acoles | but notmyname isn't watching | 17:34 |
clayg | well there *is* only *one* reclaim tombstone | 17:34 |
clayg | it's not test_hash_suffix_two_reclaim_tombstone - that's a totally different test | 17:35 |
acoles | k, silly me | 17:35 |
clayg | acoles: yeah id | 17:35 |
clayg | idk | 17:35 |
acoles | clayg: i'll submit what i have as comments, all this chat, i'm on about 8 tests from a gazillion, all good so far otherwise | 17:36 |
clayg | heh | 17:37 |
clayg | ok yeah i'm about to leave so you can make some progress | 17:37 |
clayg | i have an idea for what to do about that test but i'll wait till your done to submit everything | 17:37 |
clayg | that you SO much for your hard work reviewing | 17:37 |
* clayg remembers that one time when we were going to do fast-POST and it was going to be awesome :'( | 17:38 | |
acoles | clayg: my heart sinks as i realise the merge conflict hell i will have when i revive fast post | 17:39 |
clayg | OH NO! | 17:40 |
acoles | clayg: on query on version 25 comments too | 17:40 |
clayg | i'll help i promise | 17:40 |
clayg | acoles: i updated the gist -> https://gist.github.com/clayg/44b341b9e131289c2603 | 17:41 |
acoles | clayg: heh. you know this channel is archived now so i can hold you to that :P | 17:41 |
clayg | we'll fix ours and leave replication doing it's thing - i'll open a bug - and whoever fixes it will have a failing test already written ;) | 17:42 |
clayg | acoles: good call! | 17:42 |
clayg | acoles: i'm looking at your comments | 17:42 |
acoles | clayg: sneaky. +1. | 17:43 |
acoles | clayg: i have a hard stop today so i am going to press on with test audit. will be gone soon, you and peluse should let me know any tasks that need picking up tomorrow. | 17:44 |
clayg | test_get_hashes_creates_partition_and_pkl should cover the test at line 981? | 17:44 |
clayg | the other "dont' just do the listdir count check" comments sound good | 17:44 |
clayg | i could try and finish the audit - I see what you're doing just naming where all the tests went on the diff | 17:45 |
clayg | we could probably also use coverage | 17:45 |
*** jordanP has quit IRC | 17:48 | |
acoles | clayg: k i'll submit what i have when i leave (30 mins) | 17:51 |
acoles | notmyname: you here ? | 17:51 |
peluse | acoles, rock on man - thanks (I was out fr the last few hrs) | 17:51 |
acoles | notmyname: apologies i have a conflict with today's meeting | 17:53 |
notmyname | acoles: I'm here | 17:54 |
notmyname | acoles: ok. meeting today should be EC status update and plans for the merge to master | 17:54 |
*** mmcardle has quit IRC | 17:55 | |
acoles | notmyname: so, my 2c worth, patch 159637 code is good, only delay in my +2 on that is i'm auditing the test changes (see scrollback) | 17:55 |
patchbot | acoles: https://review.openstack.org/#/c/159637/ | 17:55 |
notmyname | ok | 17:56 |
acoles | i'll be back on that tomorrow am if its not been finished by others, | 17:56 |
acoles | then i'll move up that chain reviewing | 17:57 |
notmyname | great :-) | 17:58 |
acoles | notmyname: tdasilva and jrichli had a question about cyclic import on https://review.openstack.org/#/c/167595/ which i couldn't answer (like how to deal with it), maybe someone could advise cos we may be able to land that soon | 17:59 |
acoles | peluse: have a great trip! | 18:00 |
jrichli | brb | 18:00 |
*** jrichli has quit IRC | 18:00 | |
peluse | acoles, thanks!!! | 18:02 |
clayg | is the meeting in like 58 mins? | 18:02 |
petertr7 | I think so, according to the wiki | 18:04 |
notmyname | clayg: yes | 18:06 |
*** cutforth has joined #openstack-swift | 18:07 | |
*** jrichli has joined #openstack-swift | 18:08 | |
*** j_king has quit IRC | 18:08 | |
*** erlon has joined #openstack-swift | 18:11 | |
*** stack_ has joined #openstack-swift | 18:11 | |
*** doxavore has quit IRC | 18:20 | |
*** Gues_____ has joined #openstack-swift | 18:21 | |
*** Gues_____ has quit IRC | 18:27 | |
*** j_king has joined #openstack-swift | 18:31 | |
acoles | clayg: ok, i got most auditing done, couple i was unsure about and ran out of time. ttfn. | 18:31 |
*** acoles is now known as acoles_away | 18:32 | |
*** doxavore has joined #openstack-swift | 18:37 | |
*** jrichli has quit IRC | 18:40 | |
*** Gues_____ has joined #openstack-swift | 18:45 | |
*** jrichli has joined #openstack-swift | 18:46 | |
peluse | notmyname, any reason why we can't land patch 164108? | 18:48 |
patchbot | peluse: https://review.openstack.org/#/c/164108/ | 18:48 |
notmyname | peluse: it can land. but that will also come in with ec, right? | 18:49 |
*** ho has joined #openstack-swift | 18:49 | |
peluse | notmyname, right now its proposed to master | 18:49 |
notmyname | ya, that's fine | 18:49 |
peluse | notmyname, easy enough to abandon that one and do one against ec though | 18:49 |
notmyname | at this point, master is fine | 18:50 |
peluse | OK, it has my +2 :) | 18:50 |
peluse | zaitcev had a comment about packaing though (in the review) | 18:50 |
zaitcev | More than comment, meanwhile. I'm going to work with the maintainer of eventlet to package it in Rawhide first, see what breaks, if anything, then push it downstream. | 18:51 |
zaitcev | I think it's Padraig | 18:51 |
peluse | zaitcev, does that somehow keep us from landing the req on master? | 18:52 |
zaitcev | I don't like it, but I'm taking it as fait accompli and working to deal with the fallout | 18:53 |
notmyname | zaitcev: isn't that the point of global requirements? to give distros a set of non-conflicting stuff to use for openstack? ie won't it be with the kilo release anyway? | 18:53 |
zaitcev | I don't see any other project needing 0.16.1. We did package 0.15.2, which is the OpenStack global. | 18:54 |
notmyname | https://github.com/openstack/requirements/blob/master/global-requirements.txt#L30 | 18:55 |
notmyname | peluse: which seems to look like that patch won't land anyway | 18:55 |
zaitcev | hmm | 18:55 |
notmyname | peluse: since it doesn't exactly match | 18:56 |
notmyname | I'm not sure what the issue with 0.17.0 is | 18:56 |
zaitcev | Do we have anyone looking at switching to HTTP/2 for proxy<->a/c/o | 18:56 |
peluse | hmm, well we can tweak ours too like that one right and resubmit | 18:56 |
notmyname | zaitcev: careful. you might be volunteering for that ;-) | 18:57 |
notmyname | zaitcev: but, no | 18:57 |
zaitcev | I know Clay loves all that stuff but he's mad busy with real work | 18:57 |
*** kota_ has joined #openstack-swift | 18:59 | |
*** Bsony has joined #openstack-swift | 18:59 | |
mattoliverau | morning | 18:59 |
notmyname | hello | 18:59 |
notmyname | meeting time in #openstack-meeting | 18:59 |
*** joeljwright1 has quit IRC | 19:00 | |
*** joeljwright has joined #openstack-swift | 19:00 | |
kota_ | morning | 19:00 |
clayg | let's do this thing! | 19:01 |
openstackgerrit | paul luse proposed openstack/swift: Bump eventlet version to 0.16.1 https://review.openstack.org/164108 | 19:03 |
*** Gues_____ has quit IRC | 19:03 | |
tsg_ | peluse: thanks for 164108 update | 19:05 |
peluse | ya | 19:06 |
*** silor1 has joined #openstack-swift | 19:10 | |
*** silor has quit IRC | 19:12 | |
*** tsg_ has quit IRC | 19:12 | |
*** tsg has joined #openstack-swift | 19:17 | |
*** mahatic has quit IRC | 19:40 | |
cschwede | tdasilva: clayg: this seems to work: http://paste.openstack.org/raw/196729/ | 19:41 |
*** annegentle has quit IRC | 19:43 | |
*** joeljwright1 has joined #openstack-swift | 19:52 | |
*** joeljwright has quit IRC | 19:55 | |
notmyname | go go go | 19:55 |
notmyname | (to lunch) | 19:56 |
cschwede | peluse: enjoy your long weekend! | 19:56 |
clayg | cschwede: awesome! | 19:56 |
kota_ | (to sleep) | 19:56 |
mattoliverau | (to find some breakfast) | 19:56 |
jrichli | :-) | 19:56 |
peluse | (to get a beer) | 19:56 |
mattoliverau | lol | 19:57 |
tdasilva | jrichli: I guess we'll be here working | 19:57 |
jrichli | yup | 19:57 |
peluse | carry on :) | 19:57 |
tdasilva | jrichli: are you on the east coast? | 19:57 |
jrichli | central time | 19:57 |
mattoliverau | tdasilva, jrichli: typey typey then :P | 19:57 |
jrichli | lol | 19:57 |
tdasilva | haha | 19:57 |
*** kota_ has quit IRC | 20:01 | |
tdasilva | cschwede: thanks for that patch | 20:01 |
cschwede | tdasilva: you’re welcome! hope it helps and is fine with you | 20:01 |
clayg | ok i need to merge some stuff into multi-fi suffix hashes | 20:02 |
tdasilva | yeah, it looks good, I'll try to apply now | 20:02 |
clayg | tdasilva: thanks! | 20:02 |
jrichli | cschwede: thanks! I am glad to learn about the kosher way to deal with these things | 20:02 |
tdasilva | cschwede: just realized that policy_specified should be initialized as None instead of False, but I think everything else looks good | 20:02 |
cschwede | jrichli: you’re welcome - i don’t say this idea is kosher ;) it just avoids that cyclic import | 20:03 |
cschwede | tdasilva: oh, yes, you’re right, should be None by default | 20:03 |
*** lcurtis has joined #openstack-swift | 20:04 | |
cschwede | tdasilva: i can apply the patch if you want | 20:05 |
tdasilva | if you have it ready, please do | 20:06 |
*** Bsony has quit IRC | 20:06 | |
tdasilva | i know it is late for you and don't want to keep you | 20:06 |
*** geaaru has quit IRC | 20:06 | |
openstackgerrit | Clay Gerrard proposed openstack/swift: Multiple Fragment Archive support for suffix hashes https://review.openstack.org/159637 | 20:07 |
clayg | peluse: ok ^ i'm going to work up the chain from there and stop at recon | 20:07 |
openstackgerrit | Christian Schwede proposed openstack/swift: Select policy when running functional test https://review.openstack.org/167595 | 20:07 |
*** silor1 has quit IRC | 20:07 | |
tdasilva | clayg: thanks for fix my py27, pep mess from the proxy refactor patch | 20:07 |
tdasilva | cschwede: thanks! | 20:08 |
clayg | tdasilva: oh man it was nothing - really need to give that a review - even the commit sorta rambles on about how I was just trying it out to see if it was a good iea | 20:08 |
clayg | *idea | 20:08 |
clayg | wait - so peluse is gone already? | 20:09 |
cschwede | tdasilva: no worries, just had to do a „git add -p“ | 20:09 |
clayg | tsg: what's wrong with eventlet 0.17!? | 20:09 |
cschwede | ok, i’m out for today, have fun everyone! | 20:10 |
clayg | cschwede: thanks! | 20:10 |
*** lpabon has joined #openstack-swift | 20:10 | |
peluse | no I'm here | 20:10 |
clayg | phew | 20:10 |
peluse | flight is tomorrow am but today is our actual anniversary so I will have to bail at 6ish though | 20:10 |
clayg | peluse: so i'm just running tox on the ssync-fi-support change - it seemed to rebase cleanly on my acoles fixups to multi-fi-suffix-hashes | 20:11 |
*** annegentle has joined #openstack-swift | 20:12 | |
clayg | tdasilva: so where do you think we are on patch 164950 | 20:12 |
patchbot | clayg: https://review.openstack.org/#/c/164950/ | 20:12 |
peluse | clayg, sweet | 20:12 |
clayg | tdasilva: did you by chance get to notice what I did with the Diskfile tests with the Minxins? I was thinking about trying something similar with test.unit.proxy.controller.test_obj | 20:12 |
clayg | tdasilva: also sounds like cschwede got the functest stuff squared - that's all done and good now? | 20:13 |
tdasilva | clayg: On the replication side, we were able to keep it almost identical to master, which is good. | 20:13 |
tdasilva | yeah, i'm reviewing cschwede code now, I think I have question for him ,but need another minute | 20:14 |
tdasilva | clayg: did not look at the Diskfile tests, but that's already merged ,right...I can take a look there too | 20:14 |
clayg | peluse: ok my laptops fan just kicked on - so we must getting close to the end of the tox run :P | 20:14 |
clayg | tdasilva: don't bother - i'll see what I come up with on the ec put extraction and maybe you can look at that | 20:15 |
clayg | did you guys see https://aws.amazon.com/blogs/aws/new-cross-region-replication-for-amazon-s3/ | 20:16 |
tdasilva | clayg: ok, I was going to do another pass on the ObjectControllers to see if there was any other refactoring that could be done, but that can wait if you are going to work on it | 20:16 |
clayg | i love it when s3 releases new features and it's obvious to me how we could compose the same functionality in swift using stuff we already have or are working on - makes me feel validated that we're working on the right stuff | 20:17 |
clayg | tdasilva: yeah i'm checking it out after I type git review - I am *not* currently tracking the functet changes - or the stuff that mattoliverau said he'd review | 20:18 |
openstackgerrit | Clay Gerrard proposed openstack/swift: Add Fragment Index filter support to ssync https://review.openstack.org/165188 | 20:18 |
clayg | peluse: ^ ok rebase away - let me know if you need any help! | 20:18 |
tdasilva | clayg: yeah, almost make it looks like s3 is playing catch up with swift | 20:19 |
wer | Swift has been performing well for us... fwiw | 20:19 |
clayg | wer: THAT'S GREAT NEWS! | 20:19 |
wer | We've been using it for a about a year now I think. about ~500tb raw | 20:21 |
wer | 200MB writes and simlar reads on pretty large files. | 20:21 |
tdasilva | clayg: it's weird that they say it relies on object versioning, I wonder if it's only "one way" replication, like the other bucket is for read-only? | 20:22 |
*** Tahmina has joined #openstack-swift | 20:23 | |
peluse | great | 20:23 |
clayg | tdasilva: no i don't think it's one way | 20:23 |
clayg | tdasilva: I think the reliance on versioning is the most telling part of the implemenation | 20:23 |
clayg | tdasilva: I looked at versioning more and it's pretty clear to me their doing something like we talkd about for the shadow copy versioning scheme where you just write pointers into the bucket and store the real object somewhere else | 20:24 |
clayg | if that somewhere else was a storage policy that's doing cross regoin replication - and you container sync the manifests - you've got yourself a feature! | 20:24 |
*** jrichli has quit IRC | 20:25 | |
clayg | their version of "sync these manifests to the remote bucket" probably uses sqs - but the idea is the same - you still write the manifest in the other location, but the object - it's just in a cross region replicatied backend | 20:25 |
*** jrichli has joined #openstack-swift | 20:26 | |
tdasilva | clayg: so you can write to the "manifest" in both buckets and that wil lresult in writing to this backend object? | 20:27 |
clayg | tdasilva: i'm guessing - but we do have some pople using storage policies to set up specific geo-distributions in a multi-site swift cluster | 20:27 |
clayg | tdasilva: I think that's how versioning writes - when a container with versingin gets a PUT the real object is stored "somewhere else" and then it writes the manifest | 20:28 |
tdasilva | clayg: i see | 20:28 |
clayg | the target bucket in the remote regoin (according to the blog post) won't see the object (the manifest) until async (so that's the container sync part) | 20:28 |
clayg | but the trick is they don't have to do the object replication in nearly that real time - they only container sync manifests | 20:28 |
clayg | then it's all about how efficient they can make the backend replication between regions | 20:29 |
tdasilva | hmm...interesting | 20:29 |
clayg | i'm really curious on the initial latency of the first get to the manifest in the target bucket vs the second get - i'm guessing there's an on-demand replication of sorts where it will go ahead and store it locally in the rmeote region if it ends up pulling it over the wan on a GET | 20:29 |
clayg | i have my theories! | 20:30 |
openstackgerrit | paul luse proposed openstack/swift: Erasure Code Reconstructor https://review.openstack.org/131872 | 20:30 |
clayg | peluse: NICE! | 20:30 |
*** gyee has quit IRC | 20:30 | |
peluse | cleanup for ya somewhat | 20:30 |
peluse | :) | 20:30 |
clayg | tdasilva: so at first glance I think I can do some stuff with the ec put extraction to clean it up - so i'll be on that patch for the rest of the day i'd guess | 20:31 |
clayg | bbiab | 20:31 |
tdasilva | clayg: ok! i will finish my testing of the functest and then jump on the GET reviews | 20:31 |
openstackgerrit | paul luse proposed openstack/swift: wip: ec reconstructor probe test https://review.openstack.org/164291 | 20:33 |
*** lpabon has quit IRC | 20:37 | |
*** annegentle has quit IRC | 20:40 | |
*** Tahmina has quit IRC | 20:44 | |
*** annegentle has joined #openstack-swift | 20:50 | |
tdasilva | cschwede: still around? | 20:54 |
*** jogriffin has joined #openstack-swift | 20:55 | |
cschwede | tdasilva: kind of, semi-awake ;) | 20:55 |
jogriffin | hey guys, I need some help setting up the django swift broswer to work with an in house swift configuration | 20:56 |
jogriffin | any one have any experience with that? | 20:56 |
cschwede | jogriffin: i’m the author of django-swiftbrowser | 20:56 |
tdasilva | cschwede: lol, no worrries..just noticed that you put the container assigment inside a in-process function, so it won't get picked up from not in-process | 20:56 |
jogriffin | cschwede: oh great, i just noticed you were here :) | 20:57 |
jogriffin | cschwede: mind lending me your ear? | 20:57 |
cschwede | tdasilva: hmm, so it needs a fix or is not working at all? i wrote a quick sample (outside swift) to verify my idea, and it worked | 20:57 |
jogriffin | first question, my containers have global read set to false. can I still access them with swift browser? | 20:58 |
*** kota_ has joined #openstack-swift | 20:58 | |
cschwede | jogriffin: i’m online for a few more minutes, how can i help? | 20:58 |
cschwede | jogriffin: as long as the user is the account owner - yes | 20:58 |
cschwede | tdasilva: i’ll have another look at the patch tomorrow morning and re-verify this | 21:00 |
*** Cipher45 has quit IRC | 21:00 | |
jogriffin | cschwede: ok, makes sense. After changing the swift_auth, I used my swift_user and swift_key vars for login credentials. Unfortunately this still failed | 21:00 |
tdasilva | cschwede: sorry, didn't want to worry you, i'm looking now | 21:00 |
tdasilva | it was more of a heads up | 21:00 |
jogriffin | cschwede: what else should I take a look at modifying? is it possible to hit single container? | 21:01 |
cschwede | jogriffin: hmm, i might need to add some kidn of logging to swiftbrowser | 21:01 |
cschwede | jogriffin: for a user that is not an account owner? difficult, because the user is not allowed to list the containers | 21:02 |
*** kota_ has quit IRC | 21:02 | |
jogriffin | no for a user that is an account owner | 21:02 |
*** Cipher45 has joined #openstack-swift | 21:02 | |
*** Cipher45 has joined #openstack-swift | 21:02 | |
cschwede | tdasilva: if you find an issue with it, could you add it to the review? i will have a closer look tomorrow morning | 21:03 |
*** Bsony has joined #openstack-swift | 21:03 | |
cschwede | jogriffin: does it work with the same credentials with python-swiftclient („swift“ CLI)? | 21:03 |
cschwede | jogriffin: i mean outside of swiftbrowser? | 21:04 |
jogriffin | cschwed: with my limited knowledge of swift, I believe our containers have single users tied to them, where the user is considered the admin and the only one with r/w. | 21:04 |
jogriffin | cschwede: yeah I am able to write to my container with python-swift-client and list the objects from within the swift cli | 21:04 |
cschwede | jogriffin: are you able to list the containers? | 21:05 |
jogriffin | just the objects from within the container | 21:05 |
*** jrichli has quit IRC | 21:06 | |
jogriffin | cshwede: will you be on tomorrow? I can ping you in the am | 21:09 |
cschwede | jogriffin: that sounds like the problem. i created an issue with a possible idea, will have a closer look at it: https://github.com/cschwede/django-swiftbrowser/issues/10 | 21:10 |
cschwede | jogriffin: yes, i’ll be online tomorrow (daytime UTC) | 21:10 |
cschwede | tdasilva: actually, i forgot to update a few places in the code with my patchset :\ | 21:14 |
jogriffin | cschwede: wonderful, thanks. Ill look for you tomorrow, have a good evening | 21:16 |
cschwede | jogriffin: thanks, you too! | 21:17 |
*** jogriffin has quit IRC | 21:18 | |
*** joeljwright1 has quit IRC | 21:43 | |
*** doxavore has quit IRC | 21:43 | |
*** doxavore has joined #openstack-swift | 21:44 | |
*** annegentle has quit IRC | 21:48 | |
*** mmcardle has joined #openstack-swift | 21:56 | |
*** tsg_ has joined #openstack-swift | 22:04 | |
*** tsg has quit IRC | 22:04 | |
peluse | notmyname, patch to bump eventlet req is ready https://review.openstack.org/#/c/164108/ | 22:07 |
*** jamielennox|away is now known as jamielennox | 22:08 | |
*** mmcardle has quit IRC | 22:09 | |
*** doxavore has quit IRC | 22:10 | |
*** annegentle has joined #openstack-swift | 22:12 | |
*** stack_ has quit IRC | 22:12 | |
*** Bsony has quit IRC | 22:17 | |
*** cutforth has quit IRC | 22:21 | |
notmyname | peluse: done | 22:35 |
*** annegentle has quit IRC | 22:48 | |
*** gyee has joined #openstack-swift | 22:54 | |
openstackgerrit | Samuel Merritt proposed openstack/swift: Add functional test for multi-range GET requests. https://review.openstack.org/167828 | 22:55 |
*** tsg_ has quit IRC | 22:57 | |
*** foexle has quit IRC | 23:00 | |
*** chlong has joined #openstack-swift | 23:10 | |
*** Tyger has quit IRC | 23:14 | |
openstackgerrit | Merged openstack/swift: Bump eventlet version to 0.16.1 https://review.openstack.org/164108 | 23:16 |
*** Bsony has joined #openstack-swift | 23:23 | |
*** Bsony has quit IRC | 23:28 | |
*** lcurtis has quit IRC | 23:35 | |
*** welldannit has joined #openstack-swift | 23:37 | |
*** chlong has quit IRC | 23:37 | |
*** chlong has joined #openstack-swift | 23:37 | |
*** tsg has joined #openstack-swift | 23:42 | |
tsg | clayg: back here - saw your eventlet 0.17.0 question .. that version has a dns reverse lookup code screwed up .. https://review.openstack.org/#/c/158287/ | 23:42 |
notmyname | tsg: thansk | 23:43 |
*** annegentle has joined #openstack-swift | 23:57 |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!