Wednesday, 2025-10-22

opendevreviewMatthias Runge proposed openstack/governance master: Appoint jlarriba to Telemetry PTL  https://review.opendev.org/c/openstack/governance/+/96451607:35
opendevreviewMatthias Runge proposed openstack/governance master: Propose jlarriba to Telemetry PTL  https://review.opendev.org/c/openstack/governance/+/96451607:51
fricklerI cannot find any information about whether a PTL stepping down should be replaced via appointment or whether there should be an extra election. this is kind of vague "If an unexpected event occurs that doesn’t give you sufficient time to dedicate to the items above, it is your responsibility to step down and allow someone with more time to take over."08:15
mnasiadkaUsually it was an appointment from what I see, but I agree we should make it more detailed.08:21
opendevreviewMichal Nasiadka proposed openstack/governance master: Propose jlarriba to Telemetry PTL  https://review.opendev.org/c/openstack/governance/+/96451608:38
*** iurygregory_ is now known as iurygregory10:57
fungiin theory the tc members could decide to hold a special election and then appoint the winner, if they so chose15:24
gouthamrhttps://lists.openstack.org/archives/list/openstack-discuss@lists.openstack.org/thread/SZXCQIETIU7VP43EGYCXPWM4HIYENWX6/#SZXCQIETIU7VP43EGYCXPWM4HIYENWX616:57
gouthamrhttps://review.opendev.org/c/openstack/governance/+/62553716:58
gouthamr^ we didn't go through an "appointment" as such - just a handover like mrunge is doing16:58
gouthamr^ sorry, my context message got eaten up by bad copy paste16:59
gouthamrwe did have a similar situation in ~2018, when the mistral PTL needed to step down, and they handed the responsibilities over to a past PTL 16:59
gouthamri don't think we need a midterm election, i think it feels a bit heavy handed, we could ask the core reviewers to look at, and object to the governance change if they need to17:01
fungiagreed, unless there's more than one party expressing interest in being the interim ptl, a special election seems like overkill to me17:04
sean-k-mooneycardoe: i dont have much expirnce with it but i think grunicorn can be used instead fo uwsigi in most cases17:51
cardoeThat's my thought too.18:09
cardoeSo uWSGI advertises its on life support and not getting updates or features or eyes. The company behind it talks about other tech (like ASGI servers and Rust).18:09
cardoeI just don't want it to become another eventlet since the signaling is pretty clear.18:10
cardoeWe've just got uWSGI recommended all over the place and in use all over the place.18:10
clarkbone appraoch could be to stop recommending any specific wsgi server and just pick something for our own upstream development needs.18:16
clarkbIn theory these servers are interchangeable (though I know that there are sometimes different expectations about how the web service is loaded into the wsgi server)18:16
cardoeWell the "bug" with switching them around right now is with oslo_config and reading of config files.18:17
cardoeCause service/project (maybe got the variables wrong) are optional to oslo_config initialization18:18
cardoeSo some projects have got uWSGI specific start up to make it go back to loading /etc/$project/$project.conf18:18
cardoeIf oslo_config isn't initialized with a service/project variable it uses /proc/self/name for $project in the above.18:19
cardoeI've been trying to submit changes to not have that.18:20
clarkbsure all the more reason to stop recommending one specific thing so that we stop writing wsgi specific server code18:22
fricklerwhere is it actually getting recommended? in docs? IMO the most important step would be to make devstack support something else18:46
fricklerwhich would likely require some volunteer to actually implement that18:47
JayFspotz[m]: https://bugs.launchpad.net/ironic/+bug/2129596 :(20:01
clarkbJayF: might be worth booting the upstream image to see if it is a problem there or just in the dib updated version (dib does change some things like the filesystem from xfs to ext4 by default iirc)20:45
clarkbJayF: it almost looks like the backing disk isn't big enough for the disk image. Could it be the size of the image grew and now you're over some implicit limit?20:49
JayFyeah, like I said in the ticket I think it could be a "needs more ram", but the image itself doesn't appear to have changed size20:49
clarkbnot more ram. More disk20:49
JayFoh, so you're saying it's not corrupting on load while extracting at boot20:50
JayFthat it's an invalid image because it's hitting disk space before?20:50
clarkbas a theory anyway20:50
JayFif that was the case, then it's more likely to be a bug that we didn't go "boom" in the build process20:50
JayFthe real answer is that I don't have time to dig this deeper anytime soon, and it's another piece of evidence pointing towards centos being a less-stable platform to test against... or at least that's what it seems like now20:51
clarkbhrm though its failing on op write which in initramfs is all in memory?20:53
clarkbso ya probably unlikely that the actual disk is too small20:54
JayFyeah, if it's out of "space" it's out of ram20:54
JayFand that feels unlikely given the disk image is the same size20:54
clarkbJayF: maybe check syslog logs on the job node to see if oomkiller or similar was invoked20:55
clarkbmaybe the host just couldn't supply as much memory as you think it would20:55
clarkb(and if you were close to the edge and the distro updated pushing things over the limit...)20:56
JayFhttps://review.opendev.org/c/openstack/ironic-tempest-plugin/+/955799 is the job in question that Clif was invested in, and why I started diggin, cheking syslog20:57
JayFnothing of real interest in the syslog afaict20:57
JayFI will try a ram bump20:57
JayFpushed a ram bump and depended that patch on it20:59
clarkbJayF: are those loop devices the actual backingdisk for the baremetal node though? the ipa image is running in memory to write out the underlying disk image (cirros in this case I think) and that is the write to the loopback device that contains the disk mapping that fails?21:15
JayFit's never even booting into anaconda21:16
JayFthat driver should boot the ramdisk into anaconda, which does an install21:16
JayFwe're not even getting that far, we can't boot the kernel/ramdisk21:16
clarkbwhat is loop2 then?21:16
JayFI don't know fully. I'd buy that there's something else going on, but I haven't ID'd why it works on one and not the other21:17
clarkblooks like anaconda fetches a stage 2 image. Maybe loop2 is where it tries to put that?21:18
JayFthe job that bumps the ram seems like it may be in better shape, but it's hard to tell until it hard passes or fails so I can get to all the logs21:19
clarkbside note your logs make zuul unhappy for some reason so I can't deep link into them21:19
JayFif that's the end case, it's still not great that we spent time on this kinda mid-release churn21:19
clarkbnot sure why21:19
JayFoh?21:19
clarkbmaybe its the binary data in the log?21:20
clarkbthough that looks like utf8 maybe that I don't have codepoints for21:20
JayFtry the no-ansi log link21:20
JayFthe filtered copy21:20
JayFor download the unfiltered and pipe it thru `less -R` (note: this is dangeresque for untrusted inputs)21:20
clarkbhttps://zuul.opendev.org/t/openstack/build/143ce59633184515a372f2cf6f500f80/log/controller/logs/ironic-bm-logs/node-0_no_ansi_2025-10-20-19%3A11%3A32_log.txt21:21
JayFoh, you mean that UI. I always click raw21:21
JayFI kinda hate the built-in log viewer21:21
clarkbright you can't deep link doingthat21:21
JayFhttps://d17f8c35fcd162676eb8-c684a40c384a27e613f3f0e997584032.ssl.cf5.rackcdn.com/openstack/143ce59633184515a372f2cf6f500f80/controller/logs/ironic-bm-logs/node-0_no_ansi_2025-10-20-19%3A11%3A32_log.txt this works, yeah?21:21
clarkbI wish more people would link to it because from there you can easily go back to the change or you can go to raw etc. But if you start at raw you have to work extra hard to work back to the other zuul info and if you start at the change it super ambiguous what is actually breaking and where to look21:22
clarkbyes the raw link works but that doesn't allow deep linking to specific lines21:22
JayFoh, I get what you're saying. Yeah that's never functionality I've used before and tbh didn't know existed until now21:22
JayFusually I pull it down and grep or load the raw and ^f then copy or screenshot relevant bits to share21:23
clarkb(just as an exmaple you linked to a change above and said this is the job in question, but in reality its a change that has run many many job a few of which have failed so I have to do detective work to figure out what you're looking at)21:23
JayFwell, any the ironic-standalone-anaconda* job since 10/20 has failed21:23
JayF         ^ of21:23
clarkbyes I was able to figure out it. But its like 8 extra steps for me to see actual build logs when given a change link21:24
clarkbI think it bugs me because everyone does it. I ask for build logs and get links to changes21:24
clarkbthey aren't the same :)21:24
JayFdoing devops through a web browser straw is painful no matter what :D 21:25
clarkbthe stage2 appears to be a centos 9 stream image fwiw21:27
clarkb(so potentially multiple locations a centos stream image update could impact things?)21:29
JayFram bump is showing benefits early on22:22
JayFwe must have been right at the cusp if that took us over the edge22:22

Generated by irclog2html.py 4.0.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!