| tonyb | karolinku[m]: Thanks, I'll include that key when the hold triggers | 00:56 |
|---|---|---|
| opendevreview | Merged openstack/project-config master: update watcher repos to use -release group for stable branch https://review.opendev.org/c/openstack/project-config/+/936754 | 01:45 |
| tonyb | karolinku[m]: I've added your key to: 200.225.47.51 | 03:33 |
| tonyb | karolinku[m], jcapitao[m]: I looks like nova is configured to use TCG (emulation) but the QEMU is too old to emulate the Haswell chip. See some snippets here: https://paste.opendev.org/show/bUFd63EUw8q3bAcvfCcU/ | 03:35 |
| tonyb | karolinku[m], jcapitao[m]: I think it should be possible to reconfigure nova to use KVM instead of TCG which I think will enable all the features of Haswell | 03:36 |
| tonyb | karolinku[m], jcapitao[m]: another small bit of good news is that it's glibc causing the SIGIL so we have more debugging options | 03:37 |
| opendevreview | Tony Breeds proposed zuul/zuul-jobs master: DNM Switch to KVM https://review.opendev.org/c/zuul/zuul-jobs/+/936023 | 05:17 |
| tonyb | karolinku[m], jcapitao[m]: LIBVIRT_TYPE doesn't do what you expect in the case of zuul-jobs/ensure-devstack see https://review.opendev.org/c/zuul/zuul-jobs/+/936023/6..7/roles/ensure-devstack/templates/local.conf.j2 | 05:20 |
| tonyb | karolinku[m], jcapitao[m]: at this point I think it's be appropriate to move the discussion to #openstack-dib as we're solidly in the details of your job/config and not in general opendev policy/config | 05:22 |
| opendevreview | Tony Breeds proposed openstack/diskimage-builder master: WIP: Add support for CentOS Stream 10 https://review.opendev.org/c/openstack/diskimage-builder/+/934045 | 05:22 |
| opendevreview | Tony Breeds proposed zuul/zuul-jobs master: DNM Switch to KVM https://review.opendev.org/c/zuul/zuul-jobs/+/936023 | 06:22 |
| opendevreview | Tony Breeds proposed openstack/diskimage-builder master: WIP: Add support for CentOS Stream 10 https://review.opendev.org/c/openstack/diskimage-builder/+/934045 | 06:22 |
| tonyb | karolinku[m], jcapitao[m]: Oh neither of you are in #openstack-dib, ^^ should use ubuntu-noble as the base OS, which has a QEMU that can emulate Haswell, and also switches back to QEMU from KVM. I also removed a lot of jobs and dropped the attempts to 1 to iterate a little quicker | 06:24 |
| jcapitao[m] | tonyb: I just joined openstack-dib | 07:26 |
| opendevreview | Rafal Lewandowski proposed openstack/diskimage-builder master: Prevent from overwriting grub defaults if no variables are set https://review.opendev.org/c/openstack/diskimage-builder/+/937684 | 08:36 |
| opendevreview | Tony Breeds proposed openstack/diskimage-builder master: WIP: Add support for CentOS Stream 10 https://review.opendev.org/c/openstack/diskimage-builder/+/934045 | 10:16 |
| opendevreview | Clark Boylan proposed opendev/system-config master: Disable Zuul auto updates over the holidays https://review.opendev.org/c/opendev/system-config/+/938061 | 16:02 |
| clarkb | wow python3.12 really removed a lot of deprecated stuff | 16:33 |
| fungi | just wait for 3.13's removal of things in the stdlib | 17:13 |
| clarkb | the pbr integration test cases that test lts distro versions of setuptools and pip with pbr basically don't work at all under python3.12 | 17:13 |
| clarkb | I think I've got that largely sorted out now though and added extra jobs to cover extra stuff | 17:14 |
| clarkb | fungi: I also discovered that pbr integration testing isn't testing the change under test | 17:26 |
| clarkb | or at least not completely. The test cases are out of the change but the installed pbr within those tests cases is always the latest release... | 17:26 |
| clarkb | fixing that will have ot happen after we get the CI mostly working though | 17:27 |
| fungi | something to do with changes in how setuptools handles editable installs differently? | 17:27 |
| fungi | do the tests use editable or straight source tree to package and install that? | 17:28 |
| clarkb | fungi: nope its this: https://opendev.org/openstack/pbr/src/branch/master/tools/integration.sh#L76 its zuulv2 era find my change situation | 17:28 |
| clarkb | probably hasn't worked for like 4 or 5 years? | 17:28 |
| fungi | yeesh | 17:30 |
| clarkb | https://github.com/voxpupuli/plumbing/issues/294 | 18:35 |
| clarkb | I don't know who is still using puppet these days (zigo maybe?) | 18:35 |
| frickler | so puppet is doing chef things? tkajinam will likely be interested | 19:09 |
| clarkb | it sounds like they will stop publishing packages or builds of the software and the code itself will be thrown over the wall at undetermined intervals | 19:13 |
| clarkb | and this is driving the voxpupuli effort to consider a fork of puppet | 19:13 |
| fungi | perforce is a company known for its proprietary commercial enterprise software products (particularly their revision control software). perforce bought the company making puppet core and who held all the ip | 19:13 |
| fungi | perforce seems not to have bought puppet with a clear plan for maintaining the established open source community around it | 19:14 |
| clarkb | I guess they will make the builds/packages available behind a license agreement wall of some sort | 19:19 |
| fungi | likely | 19:19 |
| fungi | in unrelated news, the ubuntu-ports reprepro script started a vos release at 14:30:43 utc today | 19:21 |
| clarkb | just in time to publish the broken jammy kernel :) | 19:21 |
| fungi | probably going to finish sometime tomorrow, but i'll try to start another run as quickly as possible | 19:22 |
| tonyb | Ubuntu-Noble sets python to be externally managed which means that pip fails unless you run it with the --break-system-packages (or similar) flag. | 23:05 |
| tonyb | Do we have a preference for how to work with that? Should we start installing the tools we care about into a (or multiple) venvs? | 23:06 |
| clarkb | tonyb: we should probably use tool specific venvs | 23:15 |
| clarkb | I'm trying to think off the top of my head what things would hit that since most stuff is in containers now | 23:15 |
| tonyb | Well in nodepool it's podman-compose, I'm sure I hit it elsewhere also docker-compose for example but we're fixing that another way | 23:21 |
| clarkb | tonyb: I think we're trying to use `docker compose` over podman compose and docker-compose | 23:24 |
| clarkb | iirc the reason for that is it kept things more compatible with docker-compose while also being able to run on a modern system | 23:25 |
| clarkb | podman compose added more interface compatbility issues | 23:25 |
| tonyb | clarkb: I hit this, trying to get CS-10 working https://review.opendev.org/c/zuul/nodepool/+/938047 clearly that is a gross hack but it worked so I wanted to do something more correct | 23:30 |
| tonyb | *perhaps* the more correct thing is to switch to 'docker compose' (v2) on noble | 23:31 |
| tonyb | corvus: ^^ general thoughts on {podman-,docker-,docker }compose in nodepool | 23:32 |
| tonyb | but as I said I hit it elsewhere too | 23:32 |
| clarkb | oh I see. Zuul and nodepool have been transitioning to be more podman native since they don't have the same operational concerns of cnverting existing services | 23:33 |
| clarkb | for those simply changing to whatever works best (and using a venv if necessary) is probably the way to go | 23:33 |
| tonyb | Ah okay, that's doable and can be done the same way on jammy and noble | 23:37 |
| tonyb | clarkb: https://review.opendev.org/c/opendev/system-config/+/934937 is my other solution (for review) | 23:37 |
| clarkb | tonyb: I think we should avoid using the break system packages flag if we can | 23:39 |
| tonyb | looks like fungi hit it https://opendev.org/opendev/system-config/src/branch/master/docker/mailman/core/Dockerfile#L16 | 23:39 |
| clarkb | mostly because we've known that aws always a bad idea in the first place and now with containers we've got a good way to largely avoid the problem in the firstp lace | 23:39 |
| clarkb | tonyb: that Dockerfile comes fomr upstream (and I think using the flag in a container image is probably fine as its like a venv) | 23:40 |
| tonyb | clarkb: That's fair. I'll look at the right way to deal with it | 23:40 |
| clarkb | in the psat we could kinda get away with it because both sides of the problem (pypa and the distro) sort of accepted people were doing it but now both sides say don't do it and we've also got better tools to avoid doing it so seems like we should try to avoid it at least | 23:41 |
| clarkb | the noble testing I did with docker compose seems to indicate that we generally don't ahve a lot of stuff that will explode which is good | 23:41 |
| tonyb | Yeah. I'll look at it I guess the Noble updates are introducing a whole bunch of new fun | 23:43 |
| clarkb | python3.12 alone is just break everything | 23:43 |
| clarkb | I've been working on updating PBRs testing for noble and you can't run bookworm or jammy's setuptools or pip on python3.12 | 23:44 |
| clarkb | so now we have two sets of itnegration jobs for pbr one that rusn on noble that only covers python3.12 and one that runs on jammy that is able to cover libraries and tools for many other platforms | 23:44 |
| tonyb | clarkb: that's sad and frustrating did we know ahead of time how "break everything" 3.12 was going to be? | 23:45 |
| clarkb | tonyb: there were definitely indicators. | 23:46 |
| clarkb | the whole openstack off of eventlet effort has been driven by python3.12 showing up too | 23:46 |
| tonyb | clarkb: I thought is was mostly 'language features' I didn't realise it was so "ecosystem wide' | 23:47 |
| clarkb | they removed a lot of deprecated stuff from stdlib and if you didn't immediately start supporting the new thing years ago then you're broken (this is the problem with setuptools and pip) | 23:47 |
| clarkb | I think fungi is saying 3.13 does a lot more of that too so we're probably in for another round of it soon | 23:48 |
| tonyb | huzzah! | 23:48 |
| opendevreview | Tony Breeds proposed openstack/diskimage-builder master: WIP: Add support for CentOS Stream 10 https://review.opendev.org/c/openstack/diskimage-builder/+/934045 | 23:57 |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!