Disaster recovery

This document describes a feature that allows you to disable some important but computationally expensive parts of the application to relieve stress on the database during an ongoing downtime.

ci_queueing_disaster_recovery

This feature flag, if temporarily enabled, disables fair scheduling on shared runners. This can help to reduce system resource usage on the jobs/request endpoint by significantly reducing the computations being performed.

Side effects:

  • In case of a large backlog of jobs, the jobs are processed in the order they were put in the system, instead of balancing the jobs across many projects.
  • Projects which are out of quota will be run. This affects only jobs created during the last hour, as prior jobs are canceled by a periodic background worker (StuckCiJobsWorker).