Troubleshooting

Deploy fails

Where to look: the build logs attached to the failed deployment (dashboard → service → deployments). Failure notifications link straight to them.

Common causes:

• Build command errors — run the same command locally in a clean checkout
• Out-of-memory during build — reduce build parallelism or increase plan quota
• Missing variable at build time — variables are baked into the environment snapshot; check Variables & Configuration

A failed build produces no artifact; your current deployment keeps serving.

Deployment stuck or unhealthy after rollout

The rollout health gate requires ready replicas to meet your desired count before traffic moves. If the new revision never passes:

1. Check runtime logs for the new replicas
2. Verify your health check path/port — a wrong probe means no replica is ever "ready"
3. Verify the container listens on the port configured under networking

Your previous version continues serving — a failed rollout never takes traffic.

Service unreachable (502 / timeout)

Work through the chain:

1. Is at least one replica ready? (Service view shows readiness)
2. Did you change ports? The platform endpoint forwards to the configured container port — part of the deployment's runtime spec, so it needs a redeploy
3. Custom domain? Confirm DNS still points to the alias hostname and ownership verification hasn't lapsed (Domains & TLS)

Variable changed but nothing happened

Variables freeze into each deployment's environment snapshot. Changing a value only affects future deploys. Redeploy the instance to pick it up. Rollbacks restore old values too — that's by design.

Console won't open

• You need project-admin on restricted projects
• Sessions need at least one ready replica; paused instances have none
• Sessions expire: 30 minutes idle, 4 hours maximum — just open a new one

Deleted something by accident

You have 14 days. Everything deleted on Parslinks first disables (hidden, unserved) then waits before purging. Restore it from the deleted-items list — see Deleting & Restoring. After purge, artifacts may be reclaimed and restoration is no longer possible.

Scaling doesn't add replicas

Check for VOLUME_REPLICA_LOCK in instance status: an attached RWO volume pins that region to a single replica. Detach the volume (or snapshot-and-reattach elsewhere) to scale beyond one.

Quota / entitlement error when creating or scaling

Plan limits are enforced server-side. Check usage on the billing page; upgrade the plan or free resources (delete unused services/previews) to proceed.

Still stuck?

Gather this before asking support — it maps directly onto how the platform debugs:

• Workspace/project/service identifiers
• The deployment ID (dpl-… hostname contains its public ID)
• Time range and affected environment