Troubleshooting
Where to look when an installation misbehaves.
| Symptom | Look at |
|---|---|
| The install fails before anything starts | kubectl -n lembos logs job/lembos-setup — the setup job says what is missing |
helm install refuses to render | The message names the value or operator to fix |
| A pod is not ready after several minutes | kubectl -n lembos logs deploy/lembos-<service> |
| Services restart a few times on first install | Normal: they wait for PostgreSQL and Temporal to accept connections |
| Sign-in fails with a TLS or connection error | Identity's public name does not resolve, or its certificate is not trusted, inside the cluster — see Addresses |
| Requests are refused with 400 | The request arrived under a hostname the service was not given |
| The Portal disconnects every minute or so | Your ingress closes idle WebSockets — raise its timeouts |
| Everyone is signed out after a restart | Redis lost its data; give it persistence |
| Invitations never arrive | Check the SMTP relay in identity.smtp.endpoint |
| A run is stuck waiting | No Worker has room: check Worker capacity and scaling |
On a local k3s, also see the chart's troubleshooting table.