[neutron][qa] Grenade jobs in check queue
Hi, In Neutron team we are thinking about limiting a bit number of CI jobs which we are running. Currently we have e.g. 4 different grenade jobs: neutron-grenade grenade-py3 neutron-grenade-multinode neutron-grenade-dvr-multinode And jobs grenade-py3 and neutron-grenade-multinode are almost the same (same python version, same L2 agent, same legacy L3 agent, same fw driver). Only difference between those 2 jobs is that one of them is single and one is multinode (2) job. So I thought that maybe we can use only one of them and I wanted to remove grenade-py3 and left only multinode job in check queue. But grenade-py3 comes from "integrated-gate-py3” template so we probably shouldn’t remove this one. Can we run only grenade-py3 and drop neutron-grenade-multinode? Or maybe we can change grenade-py3 to be multinode job and then drop neutron-grenade-multinode? — Slawek Kaplonski Senior software engineer Red Hat
---- On Sat, 11 May 2019 16:05:05 +0900 Slawomir Kaplonski <skaplons@redhat.com> wrote ----
Hi,
In Neutron team we are thinking about limiting a bit number of CI jobs which we are running. Currently we have e.g. 4 different grenade jobs:
neutron-grenade
This should go away in the start of U cycle when we will drop the py27 support.
grenade-py3 neutron-grenade-multinode neutron-grenade-dvr-multinode
And jobs grenade-py3 and neutron-grenade-multinode are almost the same (same python version, same L2 agent, same legacy L3 agent, same fw driver). Only difference between those 2 jobs is that one of them is single and one is multinode (2) job. So I thought that maybe we can use only one of them and I wanted to remove grenade-py3 and left only multinode job in check queue. But grenade-py3 comes from "integrated-gate-py3” template so we probably shouldn’t remove this one. Can we run only grenade-py3 and drop neutron-grenade-multinode? Or maybe we can change grenade-py3 to be multinode job and then drop neutron-grenade-multinode?
Making grenade-py3 multinode jobs might cause the issue in other project gates. IMO, we should keep testing multinode and single node scenario separately. One thing you can do is not t run the multinode job on gate pipeline, running it on check pipeline only should be enough. -gmann
— Slawek Kaplonski Senior software engineer Red Hat
participants (2)
-
Ghanshyam Mann
-
Slawomir Kaplonski