| @mhuin:matrix.org | zuul-maint: could I get eyes on https://review.opendev.org/c/zuul/zuul/+/937895 ? It passed the CI in the last PS and I just added a release note. This is to add a zuul variable for jobs that is set to true if the job will trigger an autohold in case of failure. Our users need this information in order to keep some external resources created during the job up, and so far this is only obtainable with a call to Zuul's REST API. This is complex in our setup because this API is SSO-protected to abide by security standards | 00:28 |
|---|---|---|
| -@gerrit:opendev.org- Matthieu Huin https://matrix.to/#/@mhuin:matrix.org proposed: [zuul/zuul] 937895: Add "autohold_if_failure" var in job inventory https://review.opendev.org/c/zuul/zuul/+/937895 | 00:29 | |
| -@gerrit:opendev.org- Tony Breeds proposed: [zuul/nodepool] 938105: Use break-system-packages on Noble when installing podman-compose https://review.opendev.org/c/zuul/nodepool/+/938105 | 01:49 | |
| -@gerrit:opendev.org- Matthieu Huin https://matrix.to/#/@mhuin:matrix.org proposed: [zuul/zuul] 938067: QueueItem: Assign event id when logging events https://review.opendev.org/c/zuul/zuul/+/938067 | 01:49 | |
| -@gerrit:opendev.org- Tony Breeds proposed: [zuul/nodepool] 938047: Use break-system-packages on Noble when installing podman-compose https://review.opendev.org/c/zuul/nodepool/+/938047 | 01:56 | |
| @mhuin:matrix.org | One patchset on our deployment triggers this log message here: https://opendev.org/zuul/zuul/src/branch/master/zuul/model.py#L6781 - is there any tool to perform "cleaning up leaked config_errors objects in zk" as mentioned in the comment above? | 02:07 |
| -@gerrit:opendev.org- Tony Breeds proposed: [zuul/nodepool] 938047: Use break-system-packages on Noble when installing podman-compose https://review.opendev.org/c/zuul/nodepool/+/938047 | 04:29 | |
| -@gerrit:opendev.org- Zuul merged on behalf of James E. Blair https://matrix.to/#/@jim:acmegating.com: [zuul/zuul] 938077: Fix test_supercedent_github_circular_deps_closed race https://review.opendev.org/c/zuul/zuul/+/938077 | 07:32 | |
| -@gerrit:opendev.org- Matthieu Huin https://matrix.to/#/@mhuin:matrix.org proposed: [zuul/zuul] 937895: Add "autohold_if_failure" var in job inventory https://review.opendev.org/c/zuul/zuul/+/937895 | 11:46 | |
| -@gerrit:opendev.org- Matthieu Huin https://matrix.to/#/@mhuin:matrix.org proposed: [zuul/zuul] 938128: autohold REST API: add ref filter validation https://review.opendev.org/c/zuul/zuul/+/938128 | 12:04 | |
| @mnasiadka:matrix.org | Am I right that nodepool openstack driver does not support not using floating ip? | 12:44 |
| @mnasiadka:matrix.org | I see azure, gce and aws support pool.use-internal-ip - but openstack doesn't | 14:51 |
| @fungicide:matrix.org | mnasiadka: https://zuul-ci.org/docs/nodepool/latest/openstack.html#attr-providers.[openstack].pools.auto-floating-ip | 14:54 |
| @mnasiadka:matrix.org | yes, I've set that to false | 14:54 |
| @fungicide:matrix.org | openstack driver auto-detects whether you need floating ips, but you can turn that behavior off | 14:55 |
| @mnasiadka:matrix.org | And then I get this - https://paste.opendev.org/show/bRKb54XCHqziKPFVQ2Iw/ | 14:55 |
| @fungicide:matrix.org | mnasiadka: the opendev collaboratory's deployment is using it that way currently in rackspace flex | 14:55 |
| @mnasiadka:matrix.org | so you have pool.auto-floating-ip unset and it works without floating ips? | 14:56 |
| @mnasiadka:matrix.org | ah, no, it needs to be False | 14:57 |
| @fungicide:matrix.org | oh! you said "not using floating ips" | 14:57 |
| @mnasiadka:matrix.org | not using at all | 14:57 |
| @fungicide:matrix.org | in opendev, rackspace flex is our only provider where we use floating ips | 14:57 |
| @fungicide:matrix.org | all our other providers use provider networks with direct internet access instead | 14:58 |
| @fungicide:matrix.org | we don't set anything special for that | 14:58 |
| @mnasiadka:matrix.org | in my case I have a geneve network - and nodepool openstack instances and zuul/nodepool servers are in the same subnet | 14:59 |
| @fungicide:matrix.org | nodepool looks at the network attributes to figure out which available network is the public one, if more than one network is listed | 14:59 |
| @mnasiadka:matrix.org | although when I set pool.auto-floating-ip: False I get the error from the paste | 14:59 |
| @fungicide:matrix.org | sounds like nodepool thinks your network needs a fip to be publicly reachable | 15:00 |
| @fungicide:matrix.org | i guess you're trying to do this all in one internal network where the executors don't need to cross the public internet to reach your job nodes (e.g. they're not being booted in a separate public cloud provider) | 15:03 |
| @mnasiadka:matrix.org | Yes, that's the same v4 subnet, in the same cloud | 15:03 |
| @mnasiadka:matrix.org | I might need to dig in to how interface_ip is getting set - because looking at the error output - my node.interface_ip is empty here: https://opendev.org/zuul/nodepool/src/commit/8fbaf3d2954e22a72640cacc944cd3ef94e54a90/nodepool/driver/statemachine.py#L307 | 15:05 |
| @fungicide:matrix.org | from that traceback i think nodepool is unable to determine from the nova api how to reach the node in order to check its ssh public hostkeys | 15:05 |
| @fungicide:matrix.org | i don't have any examples of such a setup handy, but it should be doable. if the cloud isn't set up to make that discoverable, there's probably an override you can add to your clouds.yaml to tell nodepool which network id is reachable from your launcher | 15:07 |
| @fungicide:matrix.org | like setting that one as "public" even though it really isn't | 15:08 |
| @fungicide:matrix.org | but yeah, the default assumptions in the nodepool drivers are generally that it's connecting to things through the public internet (mainly for connecting to multiple public cloud service providers), so if you're doing it all in an isolated private network then some additional settings are needed | 15:10 |
| @mnasiadka:matrix.org | Ok - I did a quick and dirty change of https://opendev.org/zuul/nodepool/src/commit/8fbaf3d2954e22a72640cacc944cd3ef94e54a90/nodepool/driver/openstack/config.py#L92 to True, and it works now | 15:14 |
| @mnasiadka:matrix.org | Will probably think of pushing some patch after christmas | 15:14 |
| @clarkb:matrix.org | I'm like 99% certain fungi is correct and you set a clouds.yaml flag. No patch necessary | 15:16 |
| @mnasiadka:matrix.org | Well, maybe at least a patch to the docs ;-) | 15:17 |
| @jim:acmegating.com | it should be in the openstacksdk docs, first and foremost | 15:19 |
| @mnasiadka:matrix.org | https://docs.openstack.org/os-client-config/latest/user/configuration.html#per-region-settings - something like in this example? | 15:19 |
| @jim:acmegating.com | yep | 15:20 |
| @mnasiadka:matrix.org | ok, that's super dark magic style, but let me check | 15:21 |
| @clarkb:matrix.org | yes routes_externally: true is the one iirc | 15:22 |
| @jim:acmegating.com | it would be an understatement to say that the authors of shade would have greatly preferred that sort of information did not need to be provided by a user, but no facility to make that determination in some openstack clouds was available. | 15:22 |
| @jim:acmegating.com | in other words, yes, it's super dark magic and it would be great if users didn't need to know it. | 15:23 |
| @mnasiadka:matrix.org | Yeah, that works - thanks | 15:25 |
| @clarkb:matrix.org | note I don't think your choice of overlay system (geneve or vxlan etc) has an impact on this. It has to do with how the networks and subnets are configured from a neutron perspective | 15:39 |
| @clarkb:matrix.org | mnasiadka: https://opendev.org/openstack/openstacksdk/src/branch/master/openstack/cloud/_network_common.py#L122-L126 yes this seems to be the bit that auto detects it | 15:42 |
| @mnasiadka:matrix.org | But that's essentially marking it as an externally routed network, which is not really true in my case, but that works ;-) | 15:43 |
| @clarkb:matrix.org | right the problem is that "externally routable" is ambiguous because your frame of reference for external can differ. | 15:44 |
| @clarkb:matrix.org | I just wanted to clarify that geneve isn't a problem here | 15:47 |
| @mnasiadka:matrix.org | right, thanks | 15:52 |
| -@gerrit:opendev.org- Matthieu Huin https://matrix.to/#/@mhuin:matrix.org proposed on behalf of Tristan Cacqueray https://matrix.to/#/@tristanc_:matrix.org: | 16:32 | |
| - [zuul/zuul-jobs] 927600: Fix the upload-logs-s3 test playbook https://review.opendev.org/c/zuul/zuul-jobs/+/927600 | ||
| - [zuul/zuul-jobs] 927582: Update the set-zuul-log-path-fact scheme to prevent huge url https://review.opendev.org/c/zuul/zuul-jobs/+/927582 | ||
| -@gerrit:opendev.org- Joseph Kostreva proposed: [zuul/zuul] 923183: Add commentable_files variable to GerritChange https://review.opendev.org/c/zuul/zuul/+/923183 | 17:11 | |
| -@gerrit:opendev.org- James E. Blair https://matrix.to/#/@jim:acmegating.com proposed: [zuul/zuul] 938146: Limit command stdout/stderr to 1GiB https://review.opendev.org/c/zuul/zuul/+/938146 | 21:45 | |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!