| *** mikecmpbll has quit IRC | 00:10 | |
| openstackgerrit | Tim Burke proposed openstack/swift master: py3: port bufferedhttp (hopefully) https://review.openstack.org/548727 | 00:16 |
|---|---|---|
| *** gyee has quit IRC | 00:44 | |
| *** ChanServ has quit IRC | 01:00 | |
| *** ChanServ has joined #openstack-swift | 01:03 | |
| *** barjavel.freenode.net sets mode: +o ChanServ | 01:03 | |
| *** itlinux has quit IRC | 01:07 | |
| *** itlinux has joined #openstack-swift | 01:07 | |
| openstackgerrit | Merged openstack/swift master: Fix warning while building docs https://review.openstack.org/595856 | 01:38 |
| *** raginbajin has quit IRC | 01:46 | |
| *** raginbajin has joined #openstack-swift | 01:49 | |
| *** threestrands has quit IRC | 02:08 | |
| *** ianychoi has quit IRC | 02:28 | |
| *** joeljwright has quit IRC | 02:28 | |
| *** jlvillal has quit IRC | 02:28 | |
| *** rledisez has quit IRC | 02:28 | |
| *** zaitcev has quit IRC | 02:28 | |
| *** Miouge has quit IRC | 02:28 | |
| *** aluria has quit IRC | 02:28 | |
| openstackgerrit | Kota Tsuyuzaki proposed openstack/swift master: Configure diskfile per storage policy https://review.openstack.org/447129 | 02:29 |
| *** ianychoi has joined #openstack-swift | 02:30 | |
| *** joeljwright has joined #openstack-swift | 02:30 | |
| *** jlvillal has joined #openstack-swift | 02:30 | |
| *** rledisez has joined #openstack-swift | 02:30 | |
| *** zaitcev has joined #openstack-swift | 02:30 | |
| *** aluria has joined #openstack-swift | 02:30 | |
| *** Miouge has joined #openstack-swift | 02:30 | |
| *** barjavel.freenode.net sets mode: +vv joeljwright zaitcev | 02:30 | |
| *** openstack has joined #openstack-swift | 13:21 | |
| *** ChanServ sets mode: +o openstack | 13:21 | |
| *** openstackstatus has joined #openstack-swift | 13:25 | |
| *** ChanServ sets mode: +v openstackstatus | 13:25 | |
| *** itlinux-away has quit IRC | 13:53 | |
| *** itlinux has joined #openstack-swift | 14:35 | |
| *** viks__ has quit IRC | 14:38 | |
| *** links has quit IRC | 14:52 | |
| *** jlvillal is now known as jlviva-viva | 14:52 | |
| *** silor has quit IRC | 14:54 | |
| *** silor has joined #openstack-swift | 14:54 | |
| *** itlinux has quit IRC | 14:55 | |
| *** pcaruana has quit IRC | 14:59 | |
| *** itlinux has joined #openstack-swift | 15:09 | |
| *** mikecmpbll has quit IRC | 15:30 | |
| *** silor has quit IRC | 15:45 | |
| *** itlinux has quit IRC | 16:02 | |
| *** itlinux has joined #openstack-swift | 16:06 | |
| *** itlinux is now known as itlinux-away | 16:16 | |
| *** itlinux-away is now known as itlinux | 16:16 | |
| notmyname | good morning | 16:18 |
| *** gyee has joined #openstack-swift | 16:19 | |
| notmyname | clayg: timburke: tdasilva: Thoughts on potential dark data creation... | 16:27 |
| notmyname | If we have the same drives for ACO, and the drives start to fill up, I think we may have the opportunity for async pendings to get lost and create dark data. | 16:27 |
| notmyname | object PUT comes in, writes to the drive, fsyncs, then tries to update the container. Updates a different drive, gets a 507, fails, and tries to write an async pending. the current drive is now full and the async pending write fails (possibly). | 16:27 |
| notmyname | I can't see anything in the code that handles an async failing to be written. I think we simply hope that the other replicas manage to write it down. but there may be more opportunity for failure there that we realize, at least when you've got a cluster-full scenario | 16:27 |
| notmyname | am I missing something obvious? | 16:28 |
| timburke | seems to make sense... i wonder if we should start marking data/meta/ts files after either we either succeed in doing the container update or succeed in fsyncing the async pending... then have the auditor start kicking out async pendings for unmarked guys | 16:33 |
| timburke | https://review.openstack.org/#/c/212824/ would be so nice to have :-( | 16:33 |
| patchbot | patch 212824 - swift - Let developers/operators add watchers to object audit - 8h 24m 15s spent in CI | 16:33 |
| notmyname | timburke: I think we're correct in making the object server response 2xx once the object is written and not after the async is written (ie container update or async doesn't affect the object write response code). however, there is still a window (see above or similar) where the container update is lost | 16:37 |
| notmyname | or rather, doing two writes (object + async) in a single transaction that successfully rolls back if the second one fails sounds *way* harder to solve :-) | 16:38 |
| *** gkadam has quit IRC | 16:38 | |
| notmyname | cschwede: thanks for working on the py3 stuff! :-) | 16:39 |
| timburke | right -- i'm jus saying that we could switch to writing down "hey, there's no work to be done here" instead of writing down "go do this extra work later" to ensure a disk-full causes us to err on the side of doing work twice rather than not at all | 16:39 |
| *** gkadam has joined #openstack-swift | 16:52 | |
| *** silor has joined #openstack-swift | 16:57 | |
| *** silor has quit IRC | 17:20 | |
| *** itlinux is now known as itlinux-away | 17:27 | |
| *** itlinux-away is now known as itlinux | 17:29 | |
| *** itlinux is now known as itlinux-away | 17:29 | |
| *** itlinux-away is now known as itlinux | 17:30 | |
| *** itlinux is now known as itlinux-away | 17:30 | |
| *** itlinux-away has quit IRC | 17:35 | |
| notmyname | timburke: https://review.openstack.org/#/c/595316/ vs https://review.openstack.org/#/c/596210/ ? I'm not sure which should land | 18:07 |
| patchbot | patch 595316 - swift - Reference legacy-tempest-dsvm-neutron-full and neu... - 1h 44m 34s spent in CI | 18:07 |
| patchbot | patch 596210 - swift - import zuul job settings from project-config - 1h 47m 28s spent in CI | 18:07 |
| * notmyname was looking at the "mechanical changes for py3" that were proposed overnight | 18:07 | |
| timburke | some combination of both, i think | 18:08 |
| timburke | i think not running these jobs on purely requirements or setup.cfg changes is a mistake, but one that's pre-existing. i intentionally *didn't* include those when i wrote my patcch | 18:12 |
| timburke | pulling in the templates and queue and post jobs and all that it likely a good thing -- we don't really want to have anything left in project-config i think | 18:13 |
| notmyname | I'm not as concerned with the differences in excluded files. more about the other stanzas like the templates one. I'm not as sure what that means | 18:13 |
| timburke | the experimental jobs, i'm not familiar with. i find it strange that we'd have irrelevant-files for those, since they're only run on-demand anyway (by leaving a "check experimental" comment) | 18:15 |
| notmyname | what kind of stuff would we define there? like a one-off "test that these two things still work together"? | 18:16 |
| notmyname | or a performance test or soemthing like that? | 18:16 |
| timburke | i guess? *shrug* | 18:16 |
| notmyname | or is it supposed to be a staging area? like we've used non-voting jobs for | 18:16 |
| *** gkadam has quit IRC | 18:30 | |
| *** silor has joined #openstack-swift | 18:58 | |
| *** mrjk has quit IRC | 19:13 | |
| *** silor has quit IRC | 19:50 | |
| *** mrjk has joined #openstack-swift | 19:53 | |
| zaitcev | Guys | 20:04 |
| zaitcev | Anyone wants to share a room with me in Denver? | 20:04 |
| zaitcev | The best they have is $167. | 20:05 |
| * zaitcev pokes cschwede | 20:05 | |
| *** itlinux has joined #openstack-swift | 20:13 | |
| *** itlinux has quit IRC | 20:58 | |
| *** onovy has joined #openstack-swift | 20:59 | |
| openstackgerrit | Merged openstack/swift-bench master: import zuul job settings from project-config https://review.openstack.org/596212 | 21:03 |
| notmyname | timburke: I've learned a few things today... | 21:46 |
| notmyname | eventually http://logs.openstack.org/56/591156/2/check/zuul-build-dashboard/a1d106d/npm/html/jobs.html will be useful for understanding zuul job definitions | 21:46 |
| notmyname | eg http://logs.openstack.org/56/591156/2/check/zuul-build-dashboard/a1d106d/npm/html/job.html?job_name=tempest-full-py3 and keep clicking the parent link | 21:47 |
| notmyname | also, the jobs in swift that have "neutron" in them is because it's to distinguish against nova-net | 21:47 |
| notmyname | also, I sortof learned how jobs compose | 21:47 |
| notmyname | 'internally in zuul, that is a "project-pipeline job variant" which is a form of job definition. but it only applies to swift because it's attached to swift's project-pipeline configuration. but it's probably easier to think of things in programming terms and consider that more like the "invocation" of the job.' | 21:48 |
| notmyname | from corvus | 21:48 |
| notmyname | and we can think of stuff like legacy-tempest-dsvm-neutron-full in zuul_jobs and the swift repo as additive | 21:48 |
| notmyname | and eventually many, but not all, jobs are to be defined in-repo. this is close to what we thought, but not exactly | 21:49 |
| notmyname | some, I assume including legacy-tempest-dsvm-neutron-full and similar ones, will still be defined in zuul_jobs because they're cross project | 21:49 |
| notmyname | ie don't define them in each project that runs them | 21:50 |
| notmyname | https://zuul-ci.org/docs/zuul/user/config.html#configuration-loading and https://zuul-ci.org/docs/zuul/user/config.html#job have some more info on jobs | 21:51 |
| notmyname | also, lib-forward-testing-python3 is ultimately defined (with a few layers of indirection) in tempest, and it explicitly disables swift. so let's not add it to our repos | 21:51 |
| notmyname | I'm told that disabling is because swift doesn't support py3 yet, so I assume the plan is to add it later | 21:52 |
| notmyname | and also, http://codesearch.openstack.org/?q=tempest-full-py3&i=nope&files=&repos= exists and is the recommended way to find stuff like zuul job definitions that can live in any repo whatsoever | 21:53 |
| zaitcev | Hmm. | 22:18 |
| zaitcev | notmyname: do you know anyone who wants to share rooms? | 22:18 |
| *** itlinux has joined #openstack-swift | 22:19 | |
| notmyname | zaitcev: I do not know. sorry. perhaps asking on the mailing list or in the -ptg channel could find someone | 22:20 |
| timburke | notmyname: i'm skeptical of the future utility of a page that (currently; i expect the list to only grow) takes up 30 screens on my Giant Monitor of Doom™ | 22:22 |
| *** mikecmpbll has joined #openstack-swift | 22:30 | |
| *** gyee has quit IRC | 22:52 | |
| *** itlinux has quit IRC | 22:57 | |
| notmyname | timburke: ...ok.... http://logs.openstack.org/60/595260/2/gate/swift-tox-func-s3api/6a0b60f/job-output.txt.gz#_2018-08-24_22_37_22_543510 | 23:08 |
| timburke | notmyname: i've seen that sort of thing on occasion... i think it comes back to timeouts popping while talking to backend servers? | 23:10 |
| notmyname | hmm | 23:10 |
| timburke | issue a(nother) recheck when it finally reports the failure | 23:11 |
| notmyname | I did | 23:14 |
| notmyname | well, didn't wait on the failure report, but it is failed in the gate | 23:15 |
| openstackgerrit | Merged openstack/swift master: Add irrelevant-files lists for in-tree check/gate jobs https://review.openstack.org/595309 | 23:51 |
| notmyname | I was hoping the swift stuff in the gate would land ...swiftly... but alas, no such luck | 23:55 |
| notmyname | patch 595309 | 23:55 |
| patchbot | https://review.openstack.org/#/c/595309/ - swift - Add irrelevant-files lists for in-tree check/gate ... (MERGED) - 8h 11m 9s spent in CI | 23:55 |
| notmyname | yikes. 8+ hours | 23:56 |
| notmyname | have a good weekend, swift-world | 23:56 |
| *** tristanC has quit IRC | 23:56 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!