Helm installation broken

[docimin]

Issue submitter TODO list

• I've searched for an already existing issues here

Describe the bug (actual behavior)

When installing, I keep getting errors in the postgresql & web pods
Installation seems to be broken because of it, doesn't start without a fix.

Expected behavior

No response

values.yaml

Default

Helm chart version

sentry:26.15.1

Steps to reproduce

Install sentry/sentry helm chart by following the readme.

Screenshots

Logs

2025-03-11 17:23:00.192 GMT [862] STATEMENT:  SELECT "sentry_option"."id", "sentry_option"."key", "sentry_option"."last_updated", "sentry_option"."last_updated_by", "sentry_option"."value" FROM "sentry_option" WHERE "sentry_option"."key" = 'transaction-events.force-disable-internal-project' LIMIT 21

and

psycopg2.errors.UndefinedTable: UndefinedTable('relation "sentry_projectkey" does not exist\nLINE 1: ...tkey"."data", "sentry_projectkey"."use_case" FROM "sentry_pr...\n                                                             ^\n')
SQL: SELECT "sentry_projectkey"."id", "sentry_projectkey"."project_id", "sentry_projectkey"."label", "sentry_projectkey"."public_key", "sentry_projectkey"."secret_key", "sentry_projectkey"."roles", "sentry_projectkey"."status", "sentry_projectkey"."date_added", "sentry_projectkey"."rate_limit_count", "sentry_projectkey"."rate_limit_window", "sentry_projectkey"."data", "sentry_projectkey"."use_case" FROM "sentry_projectkey" WHERE ("sentry_projectkey"."project_id" = %s AND "sentry_projectkey"."roles" = ("sentry_projectkey"."roles" | %s) AND "sentry_projectkey"."use_case" = %s)

Additional context

No response

[Mokto]

This issue is stale because it has been open for 30 days with no activity.

[jmealo]

@docimin Did you find a fix for this?
@Mokto: I'm having the same issue, even though I set the timeout to 15 minutes. I never see any tables get connected with Postgres. I've triple checked my connection info.

# Disable the built-in PostgreSQL
postgresql:
  enabled: false

# Configure external PostgreSQL
externalPostgresql:
  host: staging-pg-common-pooler-rw.db.svc.cluster.local
  username: sentry
  password: <redacted>
  database: sentry
  port: 5432

# I added this because the migrations weren't running - to no effect
asHook: true

# I added this because the migrations weren't running - to no effect
hooks:
  enabled: true
  dbInit:
    enabled: true

global:
  user:
    create: true
    email: <redacted>
    password: <redacted>

ingress:
  enabled: false

service:
  annotations: {}
  externalPort: 9000
  name: sentry
  type: ClusterIP

[jmealo]

When I try to use the built-in Postgres. sentry-worker is unable to connect using the postgres user (password authentication fails).

[docimin]

Hey @jmealo, I did not and still require a fix.

[byrnedo]

Also experiencing the issue with the database not being set up correctly

[jmealo]

@byrnedo @docimin: I opened a new ticket, I thought it might be limited to Azure AKS. #1747

[jmealo]

@byrnedo @docimin: Are we trying to use the internal database or an external one? I'm happy to turn on connection logging to verify if an attempt is even being made. I triple checked that I could connect to the database using the provided credentials from inside one of the pods and it worked fine.

Did either of you try the hook bits I added in my values (posted in #1747).

[docimin]

@byrnedo @docimin: Are we trying to use the internal database or an external one? I'm happy to turn on connection logging to verify if an attempt is even being made. I triple checked that I could connect to the database using the provided credentials from inside one of the pods and it worked fine.

Did either of you try the hook bits I added in my values (posted in #1747).

Keep in mind that I tried to use the completely default helm chart, everything mostly how it was and only change things like the login.

Only thing I changed was the default user/pass for sentry itself.

[jmealo]

@docimin: Are you using managed k8s on a cloud provider (if so who?) Thanks for the quick reply.

I also cannot get the "default" chart to deploy.

I'm trying to see if there's any open PRs that could be related to this:
#1704

I saw someone else who never had the database migrations run and they had an issue with hooks running.

[docimin]

@docimin: Are you using managed k8s on a cloud provider (if so who?) Thanks for the quick reply.

I also cannot get the "default" chart to deploy.

I'm trying to see if there's any open PRs that could be related to this: #1704

I saw someone else who never had the database migrations run and they had an issue with hooks running.

I'm using Rancher with RKE2 cluster. Self hosted.

[byrnedo]

@byrnedo @docimin: Are we trying to use the internal database or an external one? I'm happy to turn on connection logging to verify if an attempt is even being made. I triple checked that I could connect to the database using the provided credentials from inside one of the pods and it worked fine.

Did either of you try the hook bits I added in my values (posted in #1747).

I'm using the defaults, so Postgres internal I guess. FWIW the first time I installed the chart in my ns it worked, then I wanted to change the ingress host, so reinstalled, and then the issue appeared. I've deleted the ns and all and still the same.

I haven't tried the hooks.

[jmealo]

Dang, I was hoping it was due to service accounts and the db init hooks, because there's two open/unmerged PRs with fixes:

• add kube service account for hook sentry-dbinit #1102
• Add serviceaccount for hooks in sentry chart #1704

I wonder if there's a way to manually run the migrations to verify that's the only issue here.

I'm going to try to reproduce on k3s locally.