stackstorm-k8s icon indicating copy to clipboard operation
stackstorm-k8s copied to clipboard

st2actionrunner reloads on adding a new pack

Open kritikash18 opened this issue 6 years ago • 3 comments

Currently, I have tried installing some custom st2packs by baking them in a docker image following https://github.com/StackStorm/st2packs-dockerfiles and it worked well. But everytime I add a new pack by updating the image, the st2actionrunner pod reloads. So if I want to be able to add packs on a regular basis, it will certainly affect the execution of the actions and workflows running at that moment. Is there another way to add packs without affecting the execution?

kritikash18 avatar Nov 08 '19 05:11 kritikash18

Originally reported in https://forum.stackstorm.com/t/stackstorm-ha-st2actionrunner-k8s-rescheduling-death-consequences/910/.

The mechanism of re-starting st2acrionrunner to pick up new packs is how the stackstorm-ha Helm chart deployment was designed and works via helm upgrades. The packs filesystem should be copied/shared via st2actionrunner container which happens on pre-init stage via sidecar containers.

There is however another edge case here which coming from the st2 core:

Yes indeed, sudden re-schedule or death of st2actionrunner leads to broken workflow execution, according to: https://github.com/StackStorm/st2/issues/4716 Ideally if st2 core could tolerate such failures with action re-schedules.

arm4b avatar Nov 08 '19 16:11 arm4b

I believe there are 2 possible ways to improve this story: handle properly st2actionrunner failures/re-schedules in st2 core or re-architect stackstorm-ha.

This could be a future enhancement.

arm4b avatar Nov 08 '19 16:11 arm4b

Hi @kritikash18 , Did you find any alternative for this. I tried installing new pack/modifying existing packs and everything actually impact running action/workflow and it is marked as 'Abandoned'.

manisha-tanwar avatar May 26 '20 23:05 manisha-tanwar