=== CentOS CI - On-boarding
build system for both
admin access to namespace), please tell us and provide a reason for them.
Currently we run tests in https://jenkins-fedora-infra.apps.ci.centos.org/job/koji/ but nodes are not updated and we've to drop all support OSes, only working node is Fedora 28 now, so we're testing our code on unsupported OS. I would like to move to CentOS instance and run tests again on current Fedoras and CentOS versions.
All the tests would be triggered by standard pagure hook.
Do you need bare-metal/vms checkout capability? (we prefer your workflow containerized) no
Resources required F32/33/rawhide, CentOS6/7/8
PVs:
Project_name: koji Project_admins: Tomáš Kopeček, Mike McLean, Yu Ming Zhu - tkopecek@redhat.com - mikem@redhat.com - yzhu@redhat.com
Metadata Update from @siddharthvipul1: - Issue assigned to siddharthvipul1 - Issue tagged with: centos-ci-infra, medium-gain, medium-trouble, namespace-request
@tkopecek I migrated fedora-infra namespace and gave you all access to it Now we have new agents (f32, f33, el7 and el8) Rawhide is ... challenging for now.
Can you login to https://jenkins-fedora-infra.apps.ocp.ci.centos.org/ and see if your ACO gives admin access to this jenkins?
@siddharthvipul1 It looks that it works, but when I try to fix agent settings to use correct one, it will fail to save (with error e.g. Logging ID=2c2f8355-5f9a-4746-962c-b4d260b3016d)
okay, so this is fixed! some permissions were messed up while migrating the pv
I just wanted to convey this as it's not too late if Koji wants it's own namespace, we can do that as well it will be a separate jenkins instance just for koji and koji related project (in case you want to separate it from fedora-infra)
It is ok to have it in same namespace. Is there some description of the agents? Our script needs to install few packages via pip (psycopg and proton-c libs) which needs to be compiled. Missing gcc is a show-stopper. Other option would be to spin docker images but I'm not sure which way/best practice is preferred in this environment.
sorry I missed this I can install needed packages for you (and you can do pip install in a venv) you can also use a container but I would recommend using podman (daemon less and less config in general) Let me know what works for you
I've rewritten it to use podman instead. Nevertheless, it means: 1) We don't need those additional nodes now, I've disabled them in project and run everything sequentially on F33 (if it is not ok, direct me to other node) 2) What about space usage? I'm pulling five images (centos7/8 f32/33/rawhide based), so it can easily eat the space. Are you cleaning podman images somehow, or should I delete them in the end of the job (space vs network usage).
I've rewritten it to use podman instead. Nevertheless, it means: 1) We don't need those additional nodes now, I've disabled them in project and run everything sequentially on F33 (if it is not ok, direct me to other node)
that's alright :) if you see trouble in that single node, I can add another one
2) What about space usage? I'm pulling five images (centos7/8 f32/33/rawhide based), so it can easily eat the space. Are you cleaning podman images somehow, or should I delete them in the end of the job (space vs network usage).
Please do clean them afterwords :)
Ack - done. Thanks for help!
hopefully migration is done and if you need any thing, do let me know and we can fix those quickly closing this ticket for now
Metadata Update from @siddharthvipul1: - Issue close_status updated to: Fixed - Issue status updated to: Closed (was: Open)
@siddharthvipul1 Ours jobs are failing now due to checkng out 'None' branch from pagure. I wanted to look if something have changed in settings and I'm not able to log in (User not authenticated at continue) Should I file some new ticket about this?
hey, it's possibly because of accounts.centos.org and fedora account system merge Are you able to log in accounts.centos.org?
or did you change your email address (because of it being different in ACO and FAS)?
Yes, I can log there (user tkopecek) + no change of e-mail address.
ping
not sure why this is happening, you are added in the openshift namespace and everything is working fine (with my test user as well) Can you try now? (also if you can give it a go in incognito/private window)
url: https://jenkins-fedora-infra.apps.ocp.ci.centos.org/
Sorry for false alarm - problem must be somewhere on my side. I've tried with clean firefox profile and it works. (probably some privacy settings/plugins in my normal profile blocks this).
Login to comment on this ticket.