customers have noticed taskrun pods getting evicted in event of:

• node drain and cluster upgrades
  • the risk here is that if a workload pod is evicted mid-cycle and scheduled on another node, then the nature of the task/pipeline can be such that it might be detrimental to run it again (read, one-time configuration, backups, deletion, infra configuration, etc)

we need to figure out how to handle these cases gracefully

Customer:

Ford