opendevreview | Boxiang Zhu proposed openstack/kolla-ansible master: Add skyline service https://review.opendev.org/c/openstack/kolla-ansible/+/828464 | 06:58 |
---|---|---|
opendevreview | Michal Nasiadka proposed openstack/kolla-ansible master: DNM: Test Cirros 0.5.2 in KVM jobs https://review.opendev.org/c/openstack/kolla-ansible/+/861324 | 09:05 |
opendevreview | Maksim Malchuk proposed openstack/kayobe stable/wallaby: Skip IP address allocation and configuration if needed https://review.opendev.org/c/openstack/kayobe/+/860652 | 09:12 |
opendevreview | Michal Nasiadka proposed openstack/kolla-ansible master: DNM: Test Cirros 0.5.2 in KVM jobs https://review.opendev.org/c/openstack/kolla-ansible/+/861324 | 09:12 |
hrw | hey | 09:18 |
hrw | mnasiadka: after playing with rabbitmq/erlang stuff I have an idea. | 09:18 |
hrw | mnasiadka: they have PPA with package for Ubuntu which they say is also fine for Debian. This would give us x86-64 and aarch64 package for both Debian and Ubuntu. | 09:19 |
mnasiadka | sounds easy to do ;) | 09:20 |
hrw | mnasiadka: their RPM package is now buildable on Fedora COPR. So I can setup repo for CS8/9 for both x86-64 and aarch64 with rebuild of their official erlang package. | 09:20 |
hrw | and we have all distro/arch combo covered | 09:21 |
hrw | and latest erlang | 09:21 |
hrw | same ver everywhere | 09:21 |
mnasiadka | nice | 09:21 |
hrw | have to check instalability and then let CI check rest | 09:21 |
opendevreview | Marcin Juszkiewicz proposed openstack/kolla master: rabbitmq: use Erlang from PPA on Debian/Ubuntu https://review.opendev.org/c/openstack/kolla/+/861096 | 09:32 |
hrw | INFO:kolla.common.utils.rabbitmq: erlang x86_64 25.1.1-1.1.el9 copr-hrw-erlang-for-rabbitmq 20 M | 09:42 |
opendevreview | Marcin Juszkiewicz proposed openstack/kolla master: rabbitmq: use Erlang from COPR on CentOS Stream/Rocky Linux https://review.opendev.org/c/openstack/kolla/+/861327 | 09:47 |
opendevreview | Maksim Malchuk proposed openstack/kayobe master: Fix an issue when 'acl' package can be forgotten https://review.opendev.org/c/openstack/kayobe/+/859842 | 09:57 |
opendevreview | Maksim Malchuk proposed openstack/kayobe master: Fix an issue when 'acl' package can be forgotten https://review.opendev.org/c/openstack/kayobe/+/859842 | 10:13 |
mnasiadka | frickler: just to make sure - we're talking about setting cpu_model=Nehalem (or something similar) in the KVM jobs? | 10:19 |
frickler | mnasiadka: yes, that was my idea. I just haven't found out yet how to do that with kolla | 10:21 |
hrw | frickler: config override for nova to set cpu_model | 10:21 |
hrw | patch to add that to globals.yml was rejected | 10:21 |
frickler | the issue is that OVH has some broken cpu flag overrides on the nodes where our CI hosts ran | 10:21 |
frickler | hrw: do you happen to have an example for that? /me is lazy today | 10:22 |
hrw | frickler: nope. never did such deploy | 10:22 |
opendevreview | Pierre Riteau proposed openstack/kolla master: Bump node_exporter to 1.4.0 https://review.opendev.org/c/openstack/kolla/+/861355 | 10:23 |
opendevreview | Michal Nasiadka proposed openstack/kolla-ansible master: CI: set cpu_model to Nehalem for kvm jobs https://review.opendev.org/c/openstack/kolla-ansible/+/861356 | 10:23 |
mnasiadka | frickler: ^^ that should do it | 10:24 |
hrw | frickler: look into tests/templates/nova-compute-overrides.j2 | 10:24 |
hrw | mnasiadka: aarch64 does not have nehalem... | 10:25 |
frickler | mnasiadka: don't you need "cpu_mode=custom" in that block, too? | 10:25 |
dougszu | is there a remove monasca patch yet? | 10:25 |
frickler | dougszu: iiuc we only deprecated it so far, removal would happen in AA | 10:26 |
opendevreview | Michal Nasiadka proposed openstack/kolla-ansible master: CI: set cpu_model to Nehalem for kvm jobs https://review.opendev.org/c/openstack/kolla-ansible/+/861356 | 10:27 |
mnasiadka | frickler: we might need to remove in Z, because we want to remove ES and put opensearch as direct replacement | 10:27 |
dougszu | yeah - so Michal's good work on OS currently supports running ES and OS at the same time | 10:28 |
dougszu | If we simplified that patch so we /replace/ ES with OS, to avoid users having to change FW etc, it would cripple Monasca | 10:28 |
frickler | that shows that I don't know anything about monasca, wasn't aware that it is related to *search | 10:29 |
dougszu | Only the logging side - that is the third option, to disable Monasca logging support in Zed | 10:30 |
dougszu | In summary options: 1) Support running ES and OS side by side in Zed release only (has some downsides) 2) Remove Monasca and ES in Zed, and add OS support 3) Remove Monasca log support only in Zed, remove ES in Zed, and add OS support | 10:33 |
dougszu | CloudKitty ES backend is another consideration | 10:33 |
opendevreview | Maksim Malchuk proposed openstack/kayobe master: Fix an issue when 'acl' package can be forgotten https://review.opendev.org/c/openstack/kayobe/+/859842 | 10:34 |
opendevreview | Pierre Riteau proposed openstack/kolla master: Bump node_exporter to 1.4.0 https://review.opendev.org/c/openstack/kolla/+/861355 | 10:35 |
opendevreview | Merged openstack/kayobe stable/victoria: Update to latest CentOS Stream 8 cloud image https://review.opendev.org/c/openstack/kayobe/+/859986 | 10:58 |
hrw | I wonder how long that 'latest' will be latest... | 11:02 |
hrw | 'Update to 2022.10 CentOS Stream 8 cloud image' probably fits better ;D | 11:03 |
hrw | (or whichever date it had) | 11:03 |
opendevreview | Michal Nasiadka proposed openstack/kolla-ansible master: CI: set cpu_model to Nehalem for kvm jobs https://review.opendev.org/c/openstack/kolla-ansible/+/861356 | 11:42 |
opendevreview | Michal Nasiadka proposed openstack/kolla master: Update support matrix https://review.opendev.org/c/openstack/kolla/+/861157 | 11:45 |
opendevreview | Marcin Juszkiewicz proposed openstack/kolla master: rabbitmq: use Erlang from COPR on CentOS Stream/Rocky Linux https://review.opendev.org/c/openstack/kolla/+/861327 | 11:47 |
hrw | dropped empty part of dockerfile | 11:47 |
opendevreview | Michal Nasiadka proposed openstack/kolla-ansible master: CI: set cpu_model to Nehalem for kvm jobs https://review.opendev.org/c/openstack/kolla-ansible/+/861356 | 11:56 |
opendevreview | Michal Nasiadka proposed openstack/kolla-ansible master: CI: set cpu_model to Nehalem for kvm jobs https://review.opendev.org/c/openstack/kolla-ansible/+/861356 | 12:00 |
mnasiadka | hrw, yoctozepto, frickler, mgoddard: Currently user survey has a question "Which images do you use?" (the binary/source and different distros like RHEL, CentOS, Ubuntu, Debian" - do we want to change it to something else? | 12:21 |
yoctozepto | mnasiadka: definitely! | 12:22 |
yoctozepto | we don't have binary any more | 12:22 |
yoctozepto | and RHEL support was dropped as well | 12:23 |
yoctozepto | I think we need to update the choce | 12:23 |
yoctozepto | and also add a question on "why" | 12:23 |
yoctozepto | with some sample answers and ability to give another explanation | 12:23 |
yoctozepto | that would come helpful | 12:23 |
yoctozepto | I am still saying that supporting more than one distro for container images is only a burden, with no actual value | 12:24 |
yoctozepto | (or very little actual value - like giving people the feeling of the same distro - in fact, people should not have to meddle inside the containers; and if they need, they should just swallow the pill - it is not like the in-container diff spectrum is so hard to learn even on distro mismatch) | 12:25 |
mnasiadka | so I'll fine tune the answers to "CentOS, Debian, Ubuntu" (no binary/source split) | 12:29 |
hrw | mnasiadka: "Which set of images do you use?", "Why this one?", "Which way of deploying Ceph you use", "IPv4 and/or IPv6?" | 12:53 |
hrw | mnasiadka: "Which OS/version on host machines?" | 12:54 |
* frickler wonders how anonymous an answer like "Contributed To: Kolla (Container deployment)|Masakari (Virtual Machine High Availability (VMHA) service)" really is ;) | 12:54 | |
hrw | mnasiadka: "Have you updated your cloud already?", "Have you skipped release?", "Which order you upgrade your cloud?" | 12:55 |
hrw | frickler: "Have you contributed to Kolla?", "Was your idea rejected?", "Was your idea reviewed?", "Was it merged?" | 13:01 |
hrw | 7 pages of open patches to k-a | 13:02 |
hrw | bbl | 13:03 |
frickler | we should do a reviewathon to at least decide which of the patches are still worth pursuing | 13:05 |
yoctozepto | frickler: ooops! | 14:27 |
yoctozepto | I like the idea of the reviewathon | 14:28 |
opendevreview | Doug Szumski proposed openstack/kolla-ansible master: Remove support for deploying OpenStack Monasca https://review.opendev.org/c/openstack/kolla-ansible/+/861392 | 14:37 |
yoctozepto | bye-bye monasca! | 14:40 |
mnasiadka | you won't be missed? :D | 14:46 |
yoctozepto | exactly | 14:57 |
* dougszu sheds tear | 14:57 | |
opendevreview | Maksim Malchuk proposed openstack/kayobe master: Fix an issue when user forgot combine custom passwords https://review.opendev.org/c/openstack/kayobe/+/861397 | 15:02 |
opendevreview | Doug Szumski proposed openstack/kolla-ansible master: Replace ElasticSearch and Kibana with OpenSearch https://review.opendev.org/c/openstack/kolla-ansible/+/856610 | 16:18 |
opendevreview | Doug Szumski proposed openstack/kolla-ansible master: Replace ElasticSearch and Kibana with OpenSearch https://review.opendev.org/c/openstack/kolla-ansible/+/856610 | 16:22 |
atmark | when upgrading from series to series, is it advisable to perform stage upgrade(e.g. --limit controller1,compute1-4 first next --limit controller2,compute5-6 etc) or just upgrade without limit? | 18:02 |
atmark | does upgrade without limit figures out everything(maintain quorum etc) ? | 18:04 |
opendevreview | Merged openstack/kayobe stable/xena: Update to latest CentOS Stream 8 cloud image https://review.opendev.org/c/openstack/kayobe/+/859983 | 18:27 |
yoctozepto | atmark: yeah, it figures out the quorum; doing all controllers at once should be fine; you can then stage parts of compute as you wish | 18:30 |
frickler | atmark: also note that in general all controllers should be upgraded before upgrading computes, both nova and neutron can get weird failing that | 18:47 |
atmark | sounds good. thanks | 18:57 |
mnasiadka | Just don’t stage computes if you’re using OVN :) | 19:02 |
atmark | follow up question, if I have second region, does the controllers in both region need to be upgraded simultaneously ? | 19:23 |
yoctozepto | atmark: no, they don't need to | 19:24 |
yoctozepto | mnasiadka: what's wrong with OVN? | 19:24 |
atmark | mnasiadka: using linuxbridge plugin but I'll make of note that. thx | 19:24 |
yoctozepto | atmark: note linuxbridge is deprecated; ovs is the most tested/stable solution | 19:25 |
atmark | deprecated on which version? | 19:25 |
atmark | i'm still on ussuri | 19:25 |
atmark | will upgrade to victoria by end of this month | 19:26 |
yoctozepto | atmark: latest, zed | 19:26 |
yoctozepto | no need to worry until then ;d | 19:26 |
atmark | is there migration from linuxbridge to ovs? | 19:26 |
yoctozepto | I think so, but probably painful; better reach out to the mailing list if you ever need guidance | 19:27 |
yoctozepto | or in other words | 19:27 |
yoctozepto | it is possible to switch at will but it may disconnect the vms until some action is taken | 19:27 |
yoctozepto | best do a testbed and test there | 19:28 |
yoctozepto | ah, and it's deprecated by the upstream, neutron, we only follow their rule :-) | 19:28 |
yoctozepto | btw, mnasiadka, fyi: https://review.opendev.org/c/openstack/project-team-guide/+/861141 | 19:29 |
atmark | gotcha. we are still far from Zed release. hopefully, by that time, someone picks up the driver and continue maintaining it. | 19:36 |
yoctozepto | atmark: I would not bet on it ;-) | 19:56 |
atmark | neutron fwaas resurrected in zed release | 20:58 |
atmark | heh | 20:58 |
*** hrww is now known as hrw | 23:08 |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!