By default, the persistentVolumeReclaimPolicy of a CephFS PV creation is set to Retain, this means that even when the user deletes this PV,
it will persist in our infrastructure in order to recover PVs unfortunately removed by mistake.
In order to delete PVs marked for deletion, we have created a cronjob in charge of deleting permanently PVs marked for deletion
in a specific period of time, this time-reclaim is set in each of the StorageClasses CIPAAS-542
Once this time-reclaim is reached, the cronJob patches the spec of the PV and sets the persistentVolumeReclaimPolicy to Delete,
this will trigger a permanently deletion of the PV.
In light of INC1973961: to mitigate the impact of something that creates and deletes PVCs in a loop, we immediately delete PVCs that were released less than the PV annotation reclaim-volumes.cern.ch/no-grace-period-if-time-since-creation-is-less-than after being created.
In order to interact with this parametrized value, the only requirement is to add the pertinent flag during the execution (e.g. -storageClassName cephfs)
Possible value:
- storageClassName: default to
cephfs, specifies the storageClassName
manila-provisioner: needed to list and annotate PVs created in the cluster and it is defined in CephFS scc deployment.
This cronjob for CephFS volumes is deployed with helm as a subchart of CephFS csi deployment.
The namespace used to be deployed is by default paas-infra-cephfs, in all the clusters.