*** jpena|off is now known as jpena | 07:20 | |
*** akekane__ is now known as abhishekk | 09:42 | |
opendevreview | Dan Smith proposed openstack/tempest master: Add a boot-snapshot-boot test https://review.opendev.org/c/openstack/tempest/+/860502 | 13:33 |
---|---|---|
*** mrjoshi_ is now known as mrjoshi | 13:40 | |
*** iurygregory_ is now known as iurygregory | 14:04 | |
dansmith | is anyone else seeing neutron failures early in job setup? | 14:15 |
dansmith | https://zuul.opendev.org/t/openstack/build/6a7714821126469abb03750db99465b1 | 14:16 |
frickler | slaweq: ralonsoh: ^^ looks like this might be related to recent stuff merged in neutron | 14:22 |
ralonsoh | let me check | 14:22 |
frickler | something about "other_config" | 14:23 |
ralonsoh | ufffff | 14:23 |
ralonsoh | ok, we need to bump the OVN version then | 14:23 |
ralonsoh | (I'm the culprit of this patch) | 14:23 |
dansmith | everything was fine last night, I make one pep8 fix to my patch this morning and .. bam, devstack setup fails! :D | 14:24 |
ralonsoh | dansmith, yes, this patch has been merged hours ago | 14:24 |
ralonsoh | I need to bump all OVN versions | 14:24 |
ralonsoh | give me some mins | 14:25 |
dansmith | ralonsoh: ack, thanks :) | 14:25 |
slaweq | thx ralonsoh | 14:26 |
frickler | I'm a bit surprised that neutron no longer seems to run the plain tempest job in their gate, is that intentional? | 14:32 |
opendevreview | Rodolfo Alonso proposed openstack/devstack master: Bump OVN version to v21.06.0 https://review.opendev.org/c/openstack/devstack/+/860577 | 14:43 |
ralonsoh | frickler, not directly, we execute it using several backends | 14:46 |
ralonsoh | this is why we created multiple child jobs | 14:46 |
ralonsoh | in any case, I've pushed https://review.opendev.org/c/openstack/devstack/+/860577 | 14:47 |
ralonsoh | by default, any job running on OVN will use this version | 14:47 |
frickler | I just commented on that. seems you broke at least nova, too | 14:47 |
ralonsoh | yeah, anyone using plain tempest on buntu 20.04 | 14:48 |
frickler | so I think it would be really good if neutron would run some unmodified job to avoid such regressions | 14:48 |
dansmith | glance | 14:59 |
dansmith | also | 14:59 |
ralonsoh | dansmith, yes, should be affected any project using this job | 15:00 |
dansmith | yeah just commenting on the "broke at least nova" :) | 15:00 |
ralonsoh | this is why I pushed this change at the beginning of this release | 15:00 |
ralonsoh | to avoid issues like this but at the end | 15:00 |
clarkb | ralonsoh: why is openstack requiring newer ovn than is supplied by the supported distro? I thought the whole point of picking distro releases to support is that we avoid this exact type of problem | 15:19 |
clarkb | the python versions, database, libvirt and so on should be driven by the supported platforms | 15:19 |
ralonsoh | clarkb, because there are new versions, installed on newer OS that are incompatible | 15:20 |
ralonsoh | clarkb, this is why, until we move to Ubuntu 22.04, I'm forcing this OVN version | 15:21 |
clarkb | ralonsoh: I understand there are newer version just like there are newer versions of python and libvirt | 15:21 |
ralonsoh | ok, I'll revert this Neutron change | 15:21 |
ralonsoh | I prefer not to start this kind of discussion | 15:21 |
clarkb | well I think it is important particularly from a qa perspecitve | 15:21 |
clarkb | openstack chooses supported platforms so that users know that if they choose to install openstack on that platform the existing software there will function and has been tested | 15:22 |
clarkb | if we're moving away from that this is fine as long as it is intentional | 15:22 |
clarkb | but it seems like it was a surprise in this case so I want to make sure that we're making decisions liek this intentionally knowing that users are potentially affected | 15:23 |
clarkb | I guess the surprise from the qa ci perspective is that devstack defaults to ovn but neutron doesn't run that job which allowed it to break? | 15:23 |
clarkb | maybe the first thing to do is ensure that neutron runs the base devstack/tempest job to avoid surprises? | 15:24 |
clarkb | (this is what frickler suggested) | 15:24 |
gmann | +1. we should run some default config job there. I was in impression that we do | 15:24 |
opendevreview | Dr. Jens Harbott proposed openstack/devstack master: DNM: Test multiple platform options https://review.opendev.org/c/openstack/devstack/+/860582 | 15:25 |
clarkb | another option instead of reverting is to support both versions of ovn at the same time | 15:26 |
clarkb | similar to how nova supports a wide range of libvirt versions or executing on python3.8 or 3.9 | 15:26 |
ralonsoh | not an option | 15:26 |
clarkb | if that isn't an option then how do we handle ovn versions differing across platforms? | 15:27 |
ralonsoh | we limit the max version | 15:27 |
ralonsoh | this feature was implemented in ovn 20.06 and removed the backwards compatibility in v22.06 | 15:27 |
ralonsoh | so for now we must install < v22.06 | 15:28 |
clarkb | (fwiw I do wish openstack would move to jammy quicker. Python3.10 in particular has shown that it is much quicker. But the decisions the project as a whole have made are to do a slower transition) | 15:29 |
frickler | o.k., reading this I'll hold my approval for 860577 | 15:30 |
frickler | we will move to 22.04 very soon, maybe most of this will be obsolete then | 15:30 |
frickler | but it would also be interesting to see what versions of OVN debian, centos and others have | 15:31 |
ralonsoh | that is defined locally in the CI jobs | 15:31 |
frickler | the platform job for debian claims that there is no OVN at all, not sure if that is still true | 15:32 |
clarkb | debian bullseye appears to have an ovn-common virtual package that is provided by openvswitch-common | 15:32 |
clarkb | but not sure if that has the correct bits | 15:32 |
frickler | ralonsoh: is there a distro that would already have v22.06? | 15:34 |
ralonsoh | no idea | 15:34 |
ralonsoh | but at least ubuntu 22.04 has a newer version | 15:35 |
ralonsoh | we need > 20.03 | 15:35 |
clarkb | jammy has 22.03 | 15:35 |
ralonsoh | perfect | 15:36 |
clarkb | and the debian ovs package doesn't appear to include any ovn binaries like the ubuntu ovn package. Pretty sure debian just doesn't do ovn out of the box. Does that mean default devstack doesn't work on debian? | 15:37 |
ralonsoh | why not? you have ML2/OVS | 15:38 |
clarkb | ralonsoh: I thought the default was ovn now though | 15:38 |
ralonsoh | yes but you can configure it with a couple of lines | 15:38 |
clarkb | right. my question is "does this work by default" | 15:38 |
ralonsoh | no idea, I know it works in Ubuntu, centos and RHEL | 15:39 |
clarkb | Looks like stackrc sets a default set of services if enabled services aren't explicitly set and that includes ovn-controller and ovn-northd. It may still work if there is a default to install from source with I guess /me looks more | 15:41 |
clarkb | OVN_BUILD_FROM_SOURCE defaults to False and I don't see any debian specific handling for it. I think that means you must set that flag on debian and devstack does not run there by default (but I've not directly tested this) | 15:42 |
opendevreview | Dr. Jens Harbott proposed openstack/devstack master: DNM: Test multiple platform options https://review.opendev.org/c/openstack/devstack/+/860582 | 15:56 |
*** soniya29|ruck is now known as soniya29 | 16:30 | |
opendevreview | James Parker proposed openstack/whitebox-tempest-plugin master: [WIP] Add vdpa tests https://review.opendev.org/c/openstack/whitebox-tempest-plugin/+/822354 | 17:50 |
*** jpena is now known as jpena|off | 18:20 |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!