Add priorityClassName to remaining pods

[marcleblanc2]

Linear issue FEIE-297: Add priorityClassName to remaining pods

• Customer's Kubernetes cluster policy blocks pods from starting if they do not have a priorityClassName in their config.
• We already had support for priorityClassName, but only for ~5 pods, need to add this for all remaining pods
• Added logic so that priorityClassName could be defined once, under the sourcegraph top level key, and / or under each pod's top-level key, which would override the config on the sourcegraph top level key, so the customer could configure:

sourcegraph:
  priorityClassName: p2

pgsql:
  priorityClassName: p1

Checklist

• Follow the manual testing process
• Update changelog
• Update Kubernetes update doc

Test plan

• Tested with Helm template
• Followed the manual testing process
• Deployed it on my test cluster, with the following override file:

priorityClasses:
- name: test
  value: 100
  preemptionPolicy: Never
  description: "test"
- name: test2
  value: 102
  preemptionPolicy: Never
  description: "test2"

sourcegraph:
  image:
    defaultTag: 6.10.3349
    useGlobalTagAsDefault: true
  priorityClassName: test

• Then re-applied, adding:

gitserver:
  priorityClassName: test2

• Both worked, output:

[2025-12-15 03:58:52] config % kubectl get pods -o custom-columns=NAME:.metadata.name,PRIORITY_CLASS:.spec.priorityClassName,PRIORITY_VALUE:.spec.priority
NAME                                         PRIORITY_CLASS   PRIORITY_VALUE
blobstore-579cbc4cb9-2gn69                   test             100
codeinsights-db-0                            test             100
codeintel-db-0                               test             100
gitserver-0                                  test2            102
gitserver-1                                  test2            102
grafana-0                                    test             100
indexed-search-0                             test             100
pgsql-0                                      test             100
precise-code-intel-worker-5b6bd8d898-9zrbg   test             100
prometheus-65468d765d-j4rgw                  test             100
redis-cache-595c746f84-2wxtf                 test             100
redis-store-5f4b87dbf4-8n24m                 test             100
searcher-0                                   test             100
sourcegraph-frontend-677d647479-77zrl        test             100
syntect-server-657b89b6f4-p59x6              test             100
worker-6d68db5b5c-twxkk                      test             100

[michaellzc]

can we DRY things up a bit using partial template?

see

deploy-sourcegraph-helm/charts/sourcegraph/templates/gitserver/gitserver.StatefulSet.yaml

Lines 100 to 102 in 604350a

	{{- include "sourcegraph.nodeSelector" (list . "gitserver" ) | trim | nindent 6 }}
	{{- include "sourcegraph.affinity" (list . "gitserver" ) | trim | nindent 6 }}
	{{- include "sourcegraph.tolerations" (list . "gitserver" ) | trim | nindent 6 }}

deploy-sourcegraph-helm/charts/sourcegraph/templates/_helpers.tpl

Lines 109 to 120 in 604350a

	{{- define "sourcegraph.nodeSelector" -}}
	{{- $top := index . 0 }}
	{{- $service := index . 1 }}
	{{- $globalNodeSelector := (index $top.Values "sourcegraph" "nodeSelector") }}
	{{- $serviceNodeSelector := (index $top.Values $service "nodeSelector") }}
	nodeSelector:
	{{- if $serviceNodeSelector }}
	{{- $serviceNodeSelector | toYaml | trim | nindent 2 }}
	{{- else if $globalNodeSelector }}
	{{- $globalNodeSelector | toYaml | trim | nindent 2 }}
	{{- end }}
	{{- end }}

[marcleblanc2]

can we DRY things up a bit using partial template?

@michaellzc Yes, I'm working on a few fixes, starting with #780, and will continue with DRYing up our templates a bit better.

This change is holding up a customer project, so I'll fast follow with the DRY updates.

[marcleblanc2]

Hey @michaellzc, I created a partial template / helper function to DRY things up a bit.

Helm templating seems to have some limitations in terms of preventing excess whitespace, so the current implementation is the cleanest way I've found so far to avoid adding whitespace when priorityClassName is not defined, without over-chomping when it is defined.

Helm whitespacing seems to be a weak point for Amp.

[marcleblanc2]

Quick and dirty bash script to find all helm charts in the repo, template them without and with --set sourcegraph.priorityClassName=test, then diff the templates without and with.
template-all-charts.sh