Thursday, 2018-04-19

*** slaweq has joined #openstack-lbaas01:20
*** kbyrne has quit IRC01:23
*** kbyrne has joined #openstack-lbaas01:23
*** slaweq has quit IRC01:25
*** AlexeyAbashkin has joined #openstack-lbaas01:39
*** AlexeyAbashkin has quit IRC01:43
*** dayou has quit IRC02:00
*** dayou has joined #openstack-lbaas02:05
*** dlundquist has joined #openstack-lbaas02:29
*** harlowja_ has quit IRC02:53
*** slaweq has joined #openstack-lbaas03:21
*** harlowja has joined #openstack-lbaas03:22
*** slaweq has quit IRC03:26
*** fnaval has joined #openstack-lbaas03:26
*** fnaval has quit IRC03:30
openstackgerritRabi Mishra proposed openstack/neutron-lbaas master: Use $NEUTRON_CORE_PLUGIN_CONF in place of $Q_PLUGIN_CONF_FILE
*** sanfern has joined #openstack-lbaas04:39
*** harlowja has quit IRC04:52
*** gans has joined #openstack-lbaas05:05
*** gans has quit IRC05:06
*** annp has quit IRC05:15
*** annp has joined #openstack-lbaas05:15
*** links has joined #openstack-lbaas05:18
*** slaweq has joined #openstack-lbaas05:22
*** slaweq has quit IRC05:27
openstackgerritGerman Eichberger proposed openstack/neutron-lbaas master: Fix proxy extension for neutron RBAC
openstackgerritGerman Eichberger proposed openstack/neutron-lbaas master: Fix proxy extension for neutron RBAC
openstackgerritGerman Eichberger proposed openstack/neutron-lbaas master: Gate API test for the lbaasv2-proxy plugin
*** pcaruana has joined #openstack-lbaas06:21
*** sapd_ has joined #openstack-lbaas06:26
*** sapd__ has quit IRC06:26
*** slaweq has joined #openstack-lbaas06:36
*** slaweq_ has joined #openstack-lbaas06:39
*** slaweq has quit IRC06:41
*** velizarx has joined #openstack-lbaas06:52
*** dims has quit IRC06:54
*** dims has joined #openstack-lbaas06:56
*** dims has quit IRC07:01
*** dims has joined #openstack-lbaas07:02
*** velizarx has quit IRC07:07
*** tesseract has joined #openstack-lbaas07:10
*** dlundquist has quit IRC07:15
*** AlexeyAbashkin has joined #openstack-lbaas07:26
*** velizarx has joined #openstack-lbaas07:27
*** dmellado has joined #openstack-lbaas07:32
*** rcernin has quit IRC07:33
*** pcaruana has quit IRC07:45
*** pcaruana has joined #openstack-lbaas07:46
*** sapd_ has quit IRC08:00
*** sapd_ has joined #openstack-lbaas08:00
openstackgerritAdam Harwell proposed openstack/neutron-lbaas master: WIP: Test l7 proxy to octavia
openstackgerritAdam Harwell proposed openstack/neutron-lbaas master: WIP: Test l7 proxy to octavia
openstackgerritAdam Harwell proposed openstack/neutron-lbaas master: WIP: Test l7 proxy to octavia
*** dayou has quit IRC08:14
*** dayou has joined #openstack-lbaas08:14
openstackgerritAdam Harwell proposed openstack/neutron-lbaas master: WIP: Test l7 proxy to octavia
*** sanfern has quit IRC11:24
*** mugsie has quit IRC11:42
*** mugsie has joined #openstack-lbaas11:42
*** mugsie has quit IRC11:42
*** mugsie has joined #openstack-lbaas11:42
*** velizarx has quit IRC11:44
*** velizarx has joined #openstack-lbaas11:48
*** atoth has joined #openstack-lbaas11:50
*** sanfern has joined #openstack-lbaas12:15
*** yamamoto_ has quit IRC12:21
*** yamamoto has joined #openstack-lbaas12:27
*** AlexeyAbashkin has quit IRC12:34
*** AlexeyAbashkin has joined #openstack-lbaas12:36
*** sanfern has quit IRC12:43
*** velizarx has quit IRC12:43
*** AlexeyAbashkin has quit IRC13:23
*** AlexeyAbashkin has joined #openstack-lbaas13:25
openstackgerritCarlos Goncalves proposed openstack/octavia master: Add sos element to Red Hat family images
*** AlexeyAbashkin has quit IRC13:52
*** AlexeyAbashkin has joined #openstack-lbaas13:57
dulekcgoncalves: Any news with DIB and PIP 10? We're hitting same issue on Kuryr stable/queens gate even though is merged and released.14:04
johnsomFYI, cutting Rocky MS1 today.14:15
cgoncalvesdulek, I haven't been following that lately so I don't know14:15
openstackgerritDoug Hellmann proposed openstack/neutron-lbaas master: uncap eventlet
*** AlexeyAbashkin has quit IRC14:23
*** AlexeyAbashkin has joined #openstack-lbaas14:25
*** AlexeyAbashkin has quit IRC14:40
*** AlexeyAbashkin has joined #openstack-lbaas14:42
*** velizarx has joined #openstack-lbaas14:51
*** AlexeyAbashkin has quit IRC14:54
*** AlexeyAbashkin has joined #openstack-lbaas14:56
*** AlexeyAbashkin has quit IRC15:04
*** AlexeyAbashkin has joined #openstack-lbaas15:08
*** yamamoto has quit IRC15:13
*** yamamoto has joined #openstack-lbaas15:14
*** yamamoto has quit IRC15:19
xgerman_johnsom: +115:20
*** yamamoto has joined #openstack-lbaas15:34
*** velizarx has quit IRC15:43
xgerman_rm_work: wonder why we kept the failover threads on a ThreadPool?
*** fnaval has joined #openstack-lbaas15:51
*** AlexeyAbashkin has quit IRC15:55
*** AlexeyAbashkin has joined #openstack-lbaas15:56
*** slaweq_ has quit IRC15:59
*** slaweq has joined #openstack-lbaas15:59
*** slaweq has quit IRC16:00
*** slaweq has joined #openstack-lbaas16:00
*** links has quit IRC16:01
*** pcaruana has quit IRC16:03
*** slaweq has quit IRC16:05
*** AlexeyAbashkin has quit IRC16:11
*** AlexeyAbashkin has joined #openstack-lbaas16:14
*** AlexeyAbashkin has quit IRC16:57
xgerman_are the gates broken?
johnsomThe actual error is above that.17:16
johnsomDoug's patch passed the gates this morning. I think the issues there are just the defects in the patch.17:18
johnsomSo, no17:19
*** yamamoto has quit IRC17:20
*** mestery has quit IRC17:29
*** SumitNaiksatam has joined #openstack-lbaas17:30
*** mestery has joined #openstack-lbaas17:31
openstackgerritMerged openstack/octavia-dashboard master: Use pool name as hint for selecting pool id
openstackgerritMerged openstack/octavia-dashboard master: Align model with v2 api
openstackgerritMerged openstack/octavia-dashboard master: Being able to change insert headers of listener
*** slaweq has joined #openstack-lbaas18:01
*** slaweq has quit IRC18:06
*** mestery has quit IRC18:07
openstackgerritGerman Eichberger proposed openstack/neutron-lbaas master: Gate API test for the lbaasv2-proxy plugin
rm_workxgerman_: i found another error that i think you hadn't found yet, in the neutron-lbaas tests18:08
*** SumitNaiksatam has quit IRC18:09
rm_workxgerman_: these changes:
rm_workthe create functions return the structure with PENDING_CREATE even though they do a wait-for-active, so when the next GET happens it compares the statuses and they're obviously wrong18:10
rm_workhmmm though it occurs to me now that this defeats the purpose of the test a bit, if we do a GET and then just do another GET immediately and compare18:11
xgerman_I think I fixed a couple of those18:11
rm_workso actually, i think i need to revert this and figure out a way to fix it better18:11
rm_workyou hadn't fixed the overall problem18:11
xgerman_well, I did this more for update18:11
rm_worki copied all your test changes18:11
rm_workthis was in addition to what you had18:11
rm_workunless you fixed it yesterday18:11
rm_workthese tests just have really dumb logic18:12
rm_worki'm not sure how we can really reconcile it18:12
xgerman_yes, I am trying to get the gates to like me again18:12
rm_worki honestly don't understand how this passes at all18:12
rm_workeven on master18:12
xgerman_we returned different stuff in n-lbaas18:12
xgerman_instead of pending we returned done18:12
xgerman_which caused all kinds of other issues18:13
xgerman_(the whole discussion of return what’s actually there instead of what the user thinks is there)18:13
rm_worklike, this18:13
rm_workerr but18:13
rm_workwe can't have returned "ACTIVE"18:14
rm_workthat'd be fumb18:14
xgerman_create_hm waits18:14
rm_workso new_hm will always be a HM in "PENDING_CREATE"18:14
rm_workyes it waits18:14
rm_workbut it returns the result of the original create18:14
rm_workso it will create the HM, save that return struct, wait for the LB to go ACTIVE, then return the original HM with "PENDING_CREATE"18:15
rm_workthen do a GET and get the HM in ACTIVE18:15
rm_workso how the frick does this ever pass?18:15
rm_workI feel like it's testing for a bug18:15
rm_worklike, "make sure that when the LB goes active, the HM status is still stuck in PENDING_CREATE"18:16
rm_workwe need to like... ignore the status difference18:16
rm_workand test it separately18:16
xgerman_n-lbaas didn’t return pending_ — I had to fix a few where it assumed ACTIVE but we return pending18:16
rm_workn-lbaas, on a *create*18:17
xgerman_n-lbaas on an update18:17
rm_workok, i'm talking only about creates18:17
xgerman_then we didn;t have all the status on all the objects in the beginning18:17
rm_workOH, do we STILL not have provisioning status on sub-objects!?18:17
rm_workthat would explain this18:17
rm_workit's not expecting to see a provisioning status at all18:18
rm_workso that's why it would pass18:18
rm_workok i'll just fix the tests to ignore that18:18
xgerman_yeah, …18:18
rm_workbecause the way i fixed this is very wrong actually.18:18
rm_workoh, also every test with "_invalid_tenant_id"18:19
rm_workoctavia ignores tenant_id on sub-objects18:20
rm_workhow were you dealing with those18:20
xgerman_yes, and I am skipping some of those tests18:20
rm_workbut not these ones18:20
rm_workagain, i copied your changes18:20
*** yamamoto has joined #openstack-lbaas18:20
rm_workso i have all of your skips18:20
xgerman_well, I didn’t work on it for a week since I thought gate/pip were broken18:20
rm_workbut you only skipped a few18:20
rm_workah ok18:20
rm_workso maybe you need to skip these too18:20
rm_worki was wondering if somehow they passed for you18:21
rm_workor if you dealt with them in your own code18:21
xgerman_also if one of your tests fail the system doesn’t roll back and subsequent tests will fila, too18:22
rm_workso annoying18:22
*** KeithMnemonic1 has quit IRC18:22
rm_worki think 75% of the fails i'm getting are due to 25% of the actual failures18:22
xgerman_ok, today I ma trying to resurrect the ProcessPool backport18:22
*** KeithMnemonic1 has joined #openstack-lbaas18:22
rm_worki still believe we need it18:23
rm_workand by "revive", you mean, +2? :P18:23
xgerman_I know we do18:23
xgerman_we revive means I back out the driver changes18:23
rm_workbecause it should still be ready to merge18:23
rm_workjust need to convince johnsom18:23
rm_workwhich driver changes?18:24
xgerman_I hole backing them out convinces him ;-) Also +2/A the otehrn changes but come back unhappy form zuul18:24
rm_workwhich driver changes18:25
rm_worki still don't know which patch(es) you're referring to18:25
xgerman_the health drivers18:25
xgerman_they used to be hardcoded18:25
rm_workoh that18:26
rm_workthat's not in this part of the patch i thought?18:26
xgerman_anyhwo, if you fix - I can +2 again18:26
xgerman_it is in the ProcessPath18:26
rm_workwe just need
rm_workand the next 318:26
rm_workoh no18:27
rm_workwe need those18:27
rm_workah which will require also lol18:28
rm_worki can get that on the end of the chain18:28
rm_worklet me clean this up18:29
xgerman_ok, that would be good. After backing out the drivers it’s just three lines18:29
xgerman_though johnsom is skeptical we need all that18:29
*** yamamoto has quit IRC18:30
rm_workwait so you want to split this patch into bits?18:31
rm_worki still don't quite understand what you mean by "backing out the drivers"18:31
rm_workin ?18:31
rm_workit's not optional18:31
rm_workthe ProcessPool thing doesn't work unless we driverize18:32
xgerman_mmh, I though it was18:33
xgerman_because drivers are a new feature (especially the LoggingDriver)18:33
rm_workbecause of the way it needs to load the function18:34
rm_workit has to be isolated18:34
xgerman_ah, gitcha18:34
xgerman_that’s why it didn’t work when just replaced Thread with Process a few months back18:35
rm_workyou can't pass class/instance functions to a ProcessPool18:35
xgerman_ok, we can have the rivers but not expose them through config?18:35
rm_workit explodes and runs nothing18:35
rm_workI mean I guess?18:35
xgerman_yeah, at least then there is no change for the user18:36
xgerman_and only under the hood18:36
rm_workthere already isn't18:36
rm_workthe config is optional as there are defaults18:36
rm_workuser can ignore the change and it will be fine18:36
rm_workdid gerrit just die?18:37
rm_workk nm it's back18:38
xgerman_well, technically we are giving them a new option but since we need that change I can be a bit more lenient18:39
rm_workLOL these are the same exact test:
*** AlexeyAbashkin has joined #openstack-lbaas18:49
*** tesseract has quit IRC18:55
*** AlexeyAbashkin has quit IRC18:59
rm_workall updated19:00
rm_workuhh, there will need to be a couple more pep8 fixes in there, i'll get to them19:01
johnsomFYI, I have a patch going for the pip issue in the stable branches of Octavia. It's just slowly working through the big gate queue today19:03
rm_worki just did a little more cherry-picking19:05
rm_workfixed up that huge Pike change for the HM stuff19:05
rm_workand snagged a couple for Queens19:05
rm_workso, this whole chain:
rm_workfor Pike19:06
rm_workand Queens is just missing these two:
johnsomrm_work Really? It doesn't do anything in that release19:07
rm_workthe update typo was in the patch before it19:08
johnsomAh, you backported the whole thing, ok. I saw german talking about pulling it apart19:10
rm_workyou can't19:10
rm_workthe reason i had to go in and driverize in that patch was because in order for the ProcessPool to load the function, it had to be split out and made into a singleton-type driver19:10
rm_workso it could be loaded post-initialization19:10
rm_workif it isn't a driver, the ProcessPool won't work19:11
rm_workand somehow those weren't in Queens either, which I didn't realize -- i thought i checked and made sure everything had made it into queens19:12
xgerman_@johnsom my thinking has evolved19:20
xgerman_we need the drivers — there is just ambiguity on how much of that is exposed to the customer…. hard code drivers… secret config…19:20
*** slaweq has joined #openstack-lbaas19:30
*** slaweq has quit IRC19:30
*** imacdonn has quit IRC19:36
*** imacdonn has joined #openstack-lbaas19:36
openstackgerritAdam Harwell proposed openstack/neutron-lbaas master: WIP: Test l7 proxy to octavia
rm_workwhew, the gate has some queues going on19:46
rm_worki wonder if it's because i submitted 10 patches at once19:47
johnsomIt's been slow all day19:47
rm_workhmm yeah seems so19:47
johnsomPart is the MS1 releases19:47
rm_worki'm just not helping19:47
rm_workk brb in a few min19:48
johnsomYeah, plus those are all going to die until we can get the pip 10 fix in19:48
*** slaweq has joined #openstack-lbaas19:54
*** harlowja has joined #openstack-lbaas20:03
johnsomThis appears to be working for our stable gates: Can the cores review so we can get it in and backported to the other branches?20:05
xgerman_still pip errors in — wonder what the other jobs did to work around that?20:26
johnsomIt's interesting that the other jobs didn't see that.20:31
johnsomI have this open against devstack that "fixes" it, but the devstack folks don't seem too happy with it:
*** atoth has quit IRC20:41
*** slaweq has quit IRC20:54
*** slaweq has joined #openstack-lbaas20:54
*** slaweq has quit IRC20:59
*** yamamoto has joined #openstack-lbaas21:49
johnsomOk Pike and Ocata: and
*** rcernin has joined #openstack-lbaas22:30
xgerman_Yeah, what’s up with that?22:34
openstackgerritAdam Harwell proposed openstack/neutron-lbaas master: WIP: Test l7 proxy to octavia
rm_workjohnsom: ummm FYI, as predicted my queens patches failed checks due to the pip issue22:51
rm_workjohnsom: but ... the whole Pike chain passed22:51
johnsomThat is odd22:52
rm_workAUGH does neutron-lbaas take "flavor_id" and we take "flavor"!? T_T22:55
rm_workjohnsom: ^^22:55
*** slaweq has joined #openstack-lbaas22:55
rm_workshould we fix our API? is this a bug?22:55
rm_workit isn't usable yet on our side, was just supposed to be compatibility22:55
johnsomha, well, it doesn't do anything really in either place so.....22:55
rm_workso maybe it should be updated?22:55
rm_workto match neutron-lbaas?22:56
rm_workor should i just add it in the compat layer22:56
johnsomWe plan it to be an ID field so, yeah, we should probably just fix22:56
*** slaweq has quit IRC23:00
openstackgerritMerged openstack/neutron-lbaas master: uncap eventlet
johnsomWow, that is so bizzare that it passed on pike23:06
johnsomSomehow it pull 8.1.1 pip23:09
johnsomIt must have been pinned in pike already???23:09
rm_workalso, your change is failing pep8 in pike23:09
rm_workbecause there's something that changed the testing23:09
rm_worki had to fix that error in my first HM patch23:10
johnsomDarn chicken/egg problem there.  I will take a look23:10
rm_workthe pep8 issues are simple23:11
rm_worklike, two bad indents23:11
rm_worknot sure what changed to make them show up23:11
johnsomThe pep8 rules changed recently I think23:11
openstackgerritAdam Harwell proposed openstack/octavia master: Correct flavor to flavor_id in LB API
johnsom pep8 fixed23:17
rm_workbut ... do we need it at all?23:18
rm_workis it fixing a non-problem?23:19
*** fnaval has quit IRC23:19
johnsomOk, I figured it out. So, pike has DIB pinned in upper constraints and that old version of dib uses the python-pip package instead of pulling in pip directly.23:20
johnsomWell, anyone building it without that old version of dib will have problems.23:21
johnsomI kind of lean towards just doing it to be safe23:21
openstackgerritAdam Harwell proposed openstack/octavia master: Add pecan middleware for neutron-lbaas compat
*** fnaval has joined #openstack-lbaas23:25
rm_workomg we run so many tests now23:27
johnsomYeah, as soon as we finish the tempest plugin we can drop the lbaas gates23:27
johnsomPlus, there is a discussion of going pure py3, but I don't expect that soon23:28
johnsomThen when neutron-lbaas is done we can ditch the v1 gates.... So, yeah, about when we retire it will be reasonable again23:29
rm_workxgerman_ / johnsom: did cascade delete not ever make it into neutron-lbaas? i thought german added it23:55
rm_workbut i can't find it23:55
johnsomI don't think so.23:57

Generated by 2.15.3 by Marius Gedminas - find it at!