| *** zzzeek has quit IRC | 00:20 | |
| *** zzzeek has joined #openstack-lbaas | 00:23 | |
| *** rcernin has quit IRC | 00:26 | |
| *** rcernin has joined #openstack-lbaas | 00:54 | |
| *** ramishra has joined #openstack-lbaas | 02:36 | |
| *** psachin has joined #openstack-lbaas | 03:51 | |
| *** vishalmanchanda has joined #openstack-lbaas | 04:39 | |
| *** psachin has quit IRC | 04:39 | |
| *** rcernin has quit IRC | 04:47 | |
| *** hongbin has quit IRC | 04:51 | |
| *** rcernin has joined #openstack-lbaas | 05:18 | |
| *** dmsimard has quit IRC | 05:55 | |
| *** dmsimard has joined #openstack-lbaas | 05:59 | |
| *** gcheresh has joined #openstack-lbaas | 06:08 | |
| *** rcernin has quit IRC | 07:26 | |
| *** rpittau|afk is now known as rpittau | 07:35 | |
| *** ataraday_ has joined #openstack-lbaas | 07:58 | |
| *** rcernin has joined #openstack-lbaas | 08:10 | |
| *** rcernin has quit IRC | 08:33 | |
| *** openstackgerrit has joined #openstack-lbaas | 10:00 | |
| openstackgerrit | lkk proposed openstack/octavia-tempest-plugin master: test quotas api https://review.opendev.org/757449 | 10:00 |
|---|---|---|
| *** gcheresh has quit IRC | 10:11 | |
| *** ataraday_ has quit IRC | 10:25 | |
| *** gcheresh has joined #openstack-lbaas | 10:31 | |
| *** ataraday_ has joined #openstack-lbaas | 10:34 | |
| *** Adri2000 has joined #openstack-lbaas | 11:24 | |
| Adri2000 | hello | 11:24 |
| Adri2000 | is it possible to create an https load-balancer with octavia if barbican is not deployed? | 11:25 |
| openstackgerrit | Ann Taraday proposed openstack/octavia master: Update amphora status on health update https://review.opendev.org/757548 | 12:13 |
| openstackgerrit | Merged openstack/octavia stable/stein: Fix python2 AttributeError with strptime https://review.opendev.org/755384 | 13:03 |
| *** TrevorV has joined #openstack-lbaas | 13:50 | |
| *** ataraday_ has quit IRC | 13:56 | |
| *** hongbin has joined #openstack-lbaas | 14:14 | |
| *** njohnston has quit IRC | 14:37 | |
| *** njohnston has joined #openstack-lbaas | 14:47 | |
| *** gcheresh has quit IRC | 15:08 | |
| johnsom | Adri2000 Without Barbican you can only do HTTPS pass through (no offload) or use a Castellan compatible secret store. | 15:25 |
| johnsom | https://docs.openstack.org/castellan/latest/ | 15:26 |
| openstackgerrit | Carlos Goncalves proposed openstack/octavia master: Fix pool delete race on load balancer cascade delete https://review.opendev.org/757604 | 15:39 |
| openstackgerrit | Carlos Goncalves proposed openstack/octavia-tempest-plugin master: Fix scenario tests issue with immutable LBs https://review.opendev.org/755777 | 15:41 |
| openstackgerrit | Carlos Goncalves proposed openstack/octavia-tempest-plugin master: Fix scenario tests issue with immutable LBs https://review.opendev.org/755777 | 15:47 |
| *** xgerman has joined #openstack-lbaas | 15:54 | |
| *** rpittau is now known as rpittau|afk | 16:01 | |
| johnsom | cgoncalves Nothing like a bug in 5 year old data model code... | 16:12 |
| cgoncalves | johnsom, please share if you know what's going on :) | 16:13 |
| johnsom | I'm guessing the pool was being used multiple times, thus the parallel flow causes an issue? I guess that isn't your test case though | 16:13 |
| cgoncalves | correct. LB + bunch of pools. no listeners, no members, nothing | 16:14 |
| johnsom | You have definitely got me a bit curious. | 16:14 |
| johnsom | My first reaction is, try block it and if it's already delete just move on. But it would be good to understand what is happening. | 16:15 |
| cgoncalves | the thing is it is not deleted by list.remove() says there's nothing to remove :S | 16:16 |
| cgoncalves | see the comments I left in the story | 16:16 |
| johnsom | But if you iterated that list after the passive fail, it's not there right? | 16:18 |
| cgoncalves | it is | 16:20 |
| cgoncalves | I LOG.error'ed the list on the raise block. see story :) | 16:21 |
| johnsom | Hmm, so I ran you test scenario and it passed without issue. | 16:55 |
| cgoncalves | run a few more times, it's not 100% reproducible | 16:55 |
| cgoncalves | you can also see the same issue in CI | 16:55 |
| cgoncalves | note both CI and local testing were on no-op drivers | 16:58 |
| johnsom | Hmm, that is interesting. I'm running live | 17:00 |
| johnsom | Yeah, mulitple runs with the full 50, no failures. So, maybe it's related to no-op going faster. | 17:07 |
| cgoncalves | +1 | 17:07 |
| johnsom | Hmm, not reproducing on noop either | 17:16 |
| johnsom | Maybe my debugging is influencing it. | 17:16 |
| johnsom | I can't seem to reproduce this. I'm running the parallel engine, noop drivers, etc. You can reproduce this locally? | 17:29 |
| johnsom | Ok, I bumped it up to 100 pools and I got one. | 17:34 |
| cgoncalves | I can reproduce locally ~25% of the time | 17:43 |
| johnsom | https://stackoverflow.com/questions/6319207/are-lists-thread-safe | 17:44 |
| johnsom | This implies list actions like remove are not thread safe in python.... | 17:45 |
| cgoncalves | I considered that, but which other thread is concurrently writing to the same list? | 17:47 |
| johnsom | The other pool deletes since it is an unordered flow | 17:47 |
| johnsom | Good news is it seems unlikely a real world use case could hit this, so it means we can take some time to address it. | 17:53 |
| cgoncalves | in order to fix the gate we need to use cascade delete in the test suite to run jobs faster, so we hit this bug | 18:05 |
| johnsom | Well, my patch actually worked, so we don't *need* to move to cascade delete | 18:06 |
| johnsom | It dropped the runtimes down to 1:30-1:45 | 18:06 |
| cgoncalves | I would like to have tests to use cascade delete, in addition to your change. it further runtimes, simplifies code and tests cascade delete that so far we have not been testing | 18:12 |
| johnsom | We have cascade delete tests | 18:12 |
| johnsom | I'm just concerned about the limits switching to cascade delete would impose on the tests. Tempest is setup to have cleanup hooks for each resource. We would be breaking that model. | 18:14 |
| johnsom | For example, if the cascade delete fails, are we leaving more resources in the cloud than we would if we relied on the tempest cleanup hooks? | 18:14 |
| cgoncalves | we wait for the LB to be deleted | 18:17 |
| cgoncalves | plus, the way tests are set now children objects are deleted also at class teardown anyway | 18:18 |
| cgoncalves | I think it is likely that cascade delete would delete more children resources even due parallel flows/tasks than deleting resources one-by-one | 18:22 |
| *** vishalmanchanda has quit IRC | 21:27 | |
| *** rcernin has joined #openstack-lbaas | 21:47 | |
| *** rcernin has quit IRC | 22:05 | |
| *** TrevorV has quit IRC | 22:29 | |
| *** rcernin has joined #openstack-lbaas | 22:29 | |
| *** rcernin has quit IRC | 22:29 | |
| *** rcernin has joined #openstack-lbaas | 22:30 | |
Generated by irclog2html.py 2.17.2 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!