plumdog
commented
1 year ago This appears to have been fixed by changing image.tag from 2.117.3 to 2.119.3.
I also tried latest in the 2.117 series, which was 2.117.9, and this also seemed to fail, so I think something is broken with workflows in 2.117 that is fixed in 2.119. But https://github.com/tryretool/retool-workflows-helm#usage says "The minimum supported image for Retool Workflows is 2.108.4", so I would expect 2.117.[latest] to work.
| From some more checking: | tag | works with workflows |
|---|---|---|
2.116.12 |
:heavy_check_mark: | |
2.117.3 |
:x: | |
2.117.9 |
:x: | |
2.119.1 |
:heavy_check_mark: | |
2.119.3 |
:heavy_check_mark: |
(it appears there are no tags in the semver range between 2.117.9 and 2.119.1 on Docker Hub:
$ curl -s "https://hub.docker.com/v2/namespaces/tryretool/repositories/backend/tags?page_size=100" | jq -r '.results[] | .name' | grep -v -e 'latest$' -e '-enterprise$' | sort -rV | grep '^2\.11[7-9]'
2.119.3
2.119.2
2.119.1
2.117.9
2.117.8
2.117.7
2.117.6
2.117.5
2.117.4
2.117.3
2.117.2)
So perhaps this fix needs to be backported to the 2.117 release line. However, from https://updates.retool.com/en/whats-new-in-self_hosted-retool-2119-45487736, none of the "fixed" items appear to have anything to do with this, so hard to tell what is going on.
Pod for the
workflow-workerdeployment is failing with the following logs:The pod then restarts as the command has exited with exit code 1, eventually going into CrashLoopBackOff.
This seems to be telling me that the workflow-worker pod can't connect to temporal. But if I port-forward to temporal from my local machine, I can talk to temporal using
tctl, so I think temporal is working. I suspect this means something is wrong in the config being passed to the workflow-worker pod.Relevant parts of my values passed to helm:
And that's deployed with chart https://charts.retool.com/retool-wf-4.11.12.tgz on Kubernetes 1.24 (AWS EKS).
All pods other than the workflow-worker are ready, and are not restarting.
What can I do to debug and fix this?