What the platform does
When a service process exits, the platform logs process exited (...); restarting in that service's log and starts it again after one second. After five crashes in five minutes it waits 15 seconds between restarts.
Triage
- Read the exit reason in the platform's log line.
out of memory (heap limit N MB)means memory (rb-006). An exit code with a stack trace on startup means the release cannot start; roll it back (rb-002). - Crashes that are minutes apart with memory climbing between them are a leak, not a bad release, unless the leak arrived with a deploy.