The resultsdb-ci-listener is a app we run in openshift. It takes messages from CI and inserts them into resultsdb's database.
Right now, changes to the app require:
Changes made in https://pagure.io/ci-resultsdb-listener/ for the app itself. Then, image changes made in https://pagure.io/kube-sig/resultsdb-ci-listener-images Then, an image built in quay.io and pushed out Finally changes in ansible to use that container (if it's not :latest).
IMHO we should move this to:
Changes made in https://pagure.io/ci-resultsdb-listener/ for the app in production and staging branches. Our openshift using s2i to see the changes and build it and deploy it.
Metadata Update from @kevin: - Issue priority set to: Waiting on Assignee (was: Needs Review) - Issue tagged with: low-gain, low-trouble, ops
Metadata Update from @patrikp: - Issue assigned to patrikp
Pagure doesn't support Openshift API for launching off builds via a commit webhook, so we're out of luck there while this remains on pagure. Maybe put this on hold until we migrate to Forgejo and revist then. I think this is supported on Forgejo.
Metadata Update from @patrikp: - Issue tagged with: blocked
Metadata Update from @patrikp: - Assignee reset
Hum. I thought we did have it working, but I am probibly mis-remembering.
In any case even if we have it only pulling when someone runs a playbook, it would still be nice to have that... because right now the image is completely out of our control. :(
So, I still think it would be good to move it, but we could of course wait more...
We are doing builds by just pointing the buildconfig to correct branch in pagure repository, but the build is executed manually by user.
The example of this is poddlers.
Just to update here, I've got s2i working locally from https://pagure.io/fork/gwmngilfen/ci-resultsdb-listener/c/77ecd7219a7121e519e1e2ac609b076f0050b085?branch=staging
I'm just working on making a PR to the Ansible repo based off the webshook2fedmsg app, and then we can test it :)
Metadata Update from @gwmngilfen: - Issue assigned to gwmngilfen
Some movement here:
That was deployed, and appears to have built correctly and is running in staging (as best I can tell, I don't really know how to verify it). Here's what I see from the pod logs in OCP:
[fedora_messaging.cli INFO] Starting consumer with resultsdb_listener.consumer:Consumer callback ... [fedora_messaging.twisted.service INFO] Authenticating with server using x509 (certfile: redacted) [fedora_messaging.twisted.protocol INFO] Successfully registered AMQP consumer Consumer(queue=resultsdb.stg_ci_listener, callback=<resultsdb_listener.consumer.Consumer object at 0x7f2fd8e23110>)
So next steps: - Can anyone recommend how to check the container is actually working? - Assuming it is, presumably the next thing is to make a "staging" branch on https://pagure.io/ci-resultsdb-listener, and then I can send a PR for the S2I changes needed.
So, it's listening for specific messages... if we can get something to generate one of those in staging we could see... should be a ci run ? but I am not sure if we have any staging ci stuff. ;(
I have added you to the fedora-infra group in pagure.io... which should give you commit there if you need it.
OK, so tricky to test if there's no CI...
If we're happy that it's working at this point, we could try rolling this out to prod? Perhaps after the freeze, just in case? I can work on getting the real repo ready and updating the Ansible code to use it, and we can merge/deploy later on, if that works for you.
Yeah, after freeze sounds good. We should be able to deploy it and find out pretty quickly if it's working as expected (and if not revert).
+1 from me to wait after freeze
Metadata Update from @phsmoura: - Issue untagged with: blocked
OK, great. I've pushed a staging branch to the real repo and opened https://pagure.io/fedora-infra/ansible/pull-request/2527 to use it. If that deploys ok I can do a PR for Prod.
Log in to comment on this ticket.