[qa][nova][cinder] ceph job has been busted on stable/pike for awhile, fix or delete?
The ceph job on stable/pike has been broken for awhile [1]. The fix to make devstack-plugin-ceph (which is branchless) work on stable/pike depends on this devstack change on stable/pike to introduce the TARGET_BRANCH variable [2]. Question is, can we move [2] forward? If not, I make a motion that we drop ceph jobs from stable/pike (and likely stable/ocata if we are still running them there) because they won't work. The ceph job is non-voting anyway so that's why it goes unnoticed for so long when it's broken. If we aren't going to fix this stuff we shouldn't be wasting CI resources running broken jobs. Did I mention that you should all get off my lawn? :) [1] https://bugs.launchpad.net/devstack-plugin-ceph/+bug/1835627 [2] https://review.opendev.org/#/c/684756/ -- Thanks, Matt
On 17-12-19 14:51:08, Matt Riedemann wrote:
The ceph job on stable/pike has been broken for awhile [1].
The fix to make devstack-plugin-ceph (which is branchless) work on stable/pike depends on this devstack change on stable/pike to introduce the TARGET_BRANCH variable [2].
Question is, can we move [2] forward? If not, I make a motion that we drop ceph jobs from stable/pike (and likely stable/ocata if we are still running them there) because they won't work.
The ceph job is non-voting anyway so that's why it goes unnoticed for so long when it's broken. If we aren't going to fix this stuff we shouldn't be wasting CI resources running broken jobs.
Did I mention that you should all get off my lawn? :)
[1] https://bugs.launchpad.net/devstack-plugin-ceph/+bug/1835627 [2] https://review.opendev.org/#/c/684756/
I'm +1 on dropping the job on stable/pike if we can't push [2] forward. This was always going to happen with EM branches anyway right? -- Lee Yarwood A5D1 9385 88CB 7E5F BE64 6618 BCA6 6E33 F672 2D76
On 12/18/2019 5:30 AM, Lee Yarwood wrote:
This was always going to happen with EM branches anyway right?
EM doesn't mean we just drop stuff if there is a reasonable way to keep it working, which we have. EM just allows us to officially no longer care if things aren't working and no one steps up to fix them. I did the latter several months ago but if I'm the only one that cares then it's not really sustainable. -- Thanks, Matt
+1 for delete -----Original Message----- From: Lee Yarwood <lyarwood@redhat.com> Sent: Wednesday, December 18, 2019 5:30 AM To: Matt Riedemann Cc: openstack-discuss@lists.openstack.org Subject: Re: [qa][nova][cinder] ceph job has been busted on stable/pike for awhile, fix or delete? On 17-12-19 14:51:08, Matt Riedemann wrote:
The ceph job on stable/pike has been broken for awhile [1].
The fix to make devstack-plugin-ceph (which is branchless) work on stable/pike depends on this devstack change on stable/pike to introduce the TARGET_BRANCH variable [2].
Question is, can we move [2] forward? If not, I make a motion that we drop ceph jobs from stable/pike (and likely stable/ocata if we are still running them there) because they won't work.
The ceph job is non-voting anyway so that's why it goes unnoticed for so long when it's broken. If we aren't going to fix this stuff we shouldn't be wasting CI resources running broken jobs.
Did I mention that you should all get off my lawn? :)
[1] https://bugs.launchpad.net/devstack-plugin-ceph/+bug/1835627 [2] https://review.opendev.org/#/c/684756/
I'm +1 on dropping the job on stable/pike if we can't push [2] forward. This was always going to happen with EM branches anyway right? -- Lee Yarwood A5D1 9385 88CB 7E5F BE64 6618 BCA6 6E33 F672 2D76
On 12/18/2019 2:20 PM, Arkady.Kanevsky@dell.com wrote:
+1 for delete
Done: https://review.opendev.org/#/c/700072/ -- Thanks, Matt
On Tue, 2019-12-17 at 14:51 -0600, Matt Riedemann wrote:
The ceph job on stable/pike has been broken for awhile [1].
The fix to make devstack-plugin-ceph (which is branchless) work on stable/pike depends on this devstack change on stable/pike to introduce the TARGET_BRANCH variable [2].
Question is, can we move [2] forward? If not, I make a motion that we drop ceph jobs from stable/pike (and likely stable/ocata if we are still running them there) because they won't work.
Drop em. If ceph was actually broken on these branches, we'd have been told by now. This will only be an issue if we backport ceph'y bugfixes that inadvertently break stuff, but who's doing that on these branches now? Stephen
The ceph job is non-voting anyway so that's why it goes unnoticed for so long when it's broken. If we aren't going to fix this stuff we shouldn't be wasting CI resources running broken jobs.
Did I mention that you should all get off my lawn? :)
[1] https://bugs.launchpad.net/devstack-plugin-ceph/+bug/1835627 [2] https://review.opendev.org/#/c/684756/
On 12/18/2019 6:02 AM, Stephen Finucane wrote:
If ceph was actually broken on these branches, we'd have been told by now. This will only be an issue if we backport ceph'y bugfixes that inadvertently break stuff, but who's doing that on these branches now?
True that I suspect most ceph-related fixes to nova would come through Red Hat people and Red Hat people don't care about anything past Queens (which is also EM). Having said that, stable/pike has been one of our more active branches lately (in nova anyway) because of gibi and elod pushing fixes there because their product at Ericsson is based on Pike. Note that stable/ocata is still running a working ceph job (I haven't dug into why it's working but the pike job isn't, maybe something to do with how the devstack-plugin-ceph repo is used in the legacy ocata job), so it'd be kind of weird to have no ceph testing on pike but have it on ocata. Maybe we should drop from both branches - it's non-voting anyway (on *all* branches). -- Thanks, Matt
On 12/17/19 3:51 PM, Matt Riedemann wrote:
The ceph job on stable/pike has been broken for awhile [1].
The fix to make devstack-plugin-ceph (which is branchless) work on stable/pike depends on this devstack change on stable/pike to introduce the TARGET_BRANCH variable [2].
Question is, can we move [2] forward? If not, I make a motion that we drop ceph jobs from stable/pike (and likely stable/ocata if we are still running them there) because they won't work.
We discussed this at today's Cinder meeting. We don't see a problem with moving [2] forward. We also don't object to simply removing non-voting jobs from stable/ocata and stable/pike. (I'm being no help here, but we did discuss it.)
The ceph job is non-voting anyway so that's why it goes unnoticed for so long when it's broken. If we aren't going to fix this stuff we shouldn't be wasting CI resources running broken jobs.
Agreed.
Did I mention that you should all get off my lawn? :)
OK, boomer. :)
[1] https://bugs.launchpad.net/devstack-plugin-ceph/+bug/1835627 [2] https://review.opendev.org/#/c/684756/
participants (5)
-
Arkady.Kanevsky＠dell.com
-
Brian Rosmaita
-
Lee Yarwood
-
Matt Riedemann
-
Stephen Finucane