concourse / concourse-bosh-release

Concourse BOSH release
Apache License 2.0
28 stars 49 forks source link

Make bbr backup more resilent #145

Closed taylorsilva closed 3 years ago

taylorsilva commented 3 years ago

fixes concourse/concourse#6588

This PR does two things:

  1. Pause the tracker and collector_pipelines components to reduce the chance of a build finishing while pg_dump is running. Sleep after pausing the components as well just to be extra safe.
  2. Call the bbr backup binary up to three times if it fails. Most likely cause of failure based on support tickets has been a build finishing and its sequence table being dropped. We think it's unlikely for this to be the cause three times in a row within 10s of each other.

CC @xtremerui