Problem:

Customers are left with lots of pipelinerun and taskrun objects after a while and have to manually delete those objects and their related resources such as PVCs used for workspaces to clean up their namespace and speed up API calls.

Goal:

Allow customers to enable and configure an auto-prune function in the operator which would automatically run the pruning for pipelnerun and taskruns and clean up old objects and their related resources such as PVCs.

A similar capability exist in the internal registry:
https://docs.openshift.com/container-platform/4.7/applications/pruning-objects.html#pruning-images_pruning-objects

Acceptance Criteria

• Admin can configure the operator to automatically prune the pipelinerun with the following configurations
  • Keep last N pipelineruns for each pipeline
  • Keep pipelineruns younger than D days for each pipeline
  • Keep pipelineruns and taskruns that are younger than D days
  • Upgrade behavior: on upgrade the pruning enable/disable should honor already existing pruneSpec in TektonConfig CR. (ie, no automatically enabling autoPruning on upgrade to 1.6)
  • On fresh install keep autoPruning enabled by default.

Note:

We tried two different approaches to achieve pruning for each pipeline resources 

• Keep last N pipelineruns for each pipeline
• Keep pipelineruns younger than D days for each pipeline

But we analyzed that this is not scalable. And we need to think of other ways to achieve this. 

Different approaches we tried can be found here. 

Epic to take it forward SRVKP-1702