Bernedotcom2312 opened a new issue, #42997:
URL: https://github.com/apache/superset/issues/42997
### Bug description
The Helm chart ships **both** `minAvailable: 1` and `maxUnavailable: 1` as
defaults in every
`podDisruptionBudget` block, while the corresponding templates `fail` the
render when both are set.
Since Helm deep-merges user values *on top of* the chart defaults, enabling
a PDB the documented
minimal way `enabled: true` inherits the chart's own default
`maxUnavailable: 1` next to the
default `minAvailable: 1`, trips the guard, and aborts **the entire `helm
template` / `install` /
`upgrade`**, not just the PDB resource.
In short: the chart's committed defaults violate the invariant its own
templates enforce. There is
no way to set `enabled: true` alone without tripping it, and the requirement
to null out one of the
two fields isn't documented in the values comments or the README.
The guard in `helm/superset/templates/pdb.yaml`:
```gotemplate
{{- with .Values.supersetNode.podDisruptionBudget }}
{{- if .enabled -}}
{{- if and .minAvailable .maxUnavailable }}
{{- fail "Only one of minAvailable or maxUnavailable should be set" }}
{{- end}}
```
The conflicting defaults in `helm/superset/values.yaml` (lines ~408-414,
repeated for each component):
```yaml
podDisruptionBudget:
enabled: false
minAvailable: 1
maxUnavailable: 1
```
### Reproduction steps
```bash
helm repo add superset https://apache.github.io/superset
helm repo update
cat > pdb-repro.yaml <<'EOF'
supersetNode:
podDisruptionBudget:
enabled: true
supersetWorker:
podDisruptionBudget:
enabled: true
EOF
helm template superset superset/superset --version 0.22.4 -f pdb-repro.yaml
```
**Expected:** a `PodDisruptionBudget` for `supersetNode` and
`supersetWorker`, each with
`minAvailable: 1` (the chart's own default).
**Actual:**
```
Error: execution error at (superset/templates/pdb.yaml:22:4):
Only one of minAvailable or maxUnavailable should be set
```
Every other resource in the chart fails to render too, not just the PDB.
Setting a value explicitly fails identically, because the default
`maxUnavailable: 1` is still
merged underneath:
```yaml
supersetNode:
podDisruptionBudget:
enabled: true
minAvailable: 2
```
### Screenshots/recordings
_No response_
### Superset version
master / latest-dev
### Python version
3.11
### Node version
16
### Browser
Chrome
### Additional context
Not a Superset application bug it is in the Helm chart, so no Python logs or
stacktrace exist.
**Affected chart versions:** `0.22.4`. Introduced in #27163
(Feb 2024) and unchanged since. Confirmed still broken on `master` by
rendering the published
chart, not only by reading the templates.
**Affected values blocks** all six carry the same conflicting defaults:
`supersetNode`, `supersetWorker`, `supersetCeleryBeat`,
`supersetCeleryFlower`,
`supersetWebsockets`, `supersetMcp`.
Templates: `pdb.yaml`, `pdb-worker.yaml`, `pdb-beat.yaml`,
`pdb-flower.yaml`, `pdb-ws.yaml`,
`pdb-mcp.yaml`.
**Workaround** for chart consumers: explicitly null the field you don't
want, e.g.
```yaml
supersetNode:
podDisruptionBudget:
enabled: true
minAvailable: 2
maxUnavailable: null
```
With GitOps tooling this is more awkward: Kustomize strategic-merge treats
`null` in a values patch
as key deletion, which would also drop the key you are trying to set, so a
JSON6902 patch is needed
to route around the chart bug.
**Mirror case:** setting only `maxUnavailable` also fails, against the
default `minAvailable: 1`,
and the error message names neither the component nor the remedy.
**Proposed fix:** default `maxUnavailable` to null in all six blocks,
keeping the `fail` guard for
users who genuinely set both. Non-breaking : anyone overriding
`maxUnavailable` explicitly is
unaffected.
A PR implementing this is open: apache/superset#42995
### Checklist
- [x] I have searched Superset docs and Slack and didn't find a solution to
my problem.
- [x] I have searched the GitHub issue tracker and didn't find a similar bug
report.
- [x] I have checked Superset's logs for errors and if I found a relevant
Python stacktrace, I included it here as text in the "additional context"
section.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
