On 2022-11-29 15:55, Aurelien Bompard wrote:
Well, we can actually do persistent storage in the ocp4 cluster. ;)
Oh, that's interesting! Are we using it already in one of our ansible-deployed apps?
You can take the meetbot app as an example: https://pagure.io/fedora-infra/ansible/blob/main/f/roles/openshift-apps/mote...
There are 4 types of persistent storage available you can use: - NFS (no storageClass): poor performance, need to be provisioned beforehand on Netapp. Can be shared by multiple pods and outside of Openshift. - RBD (storageClass: ocs-storagecluster-ceph-rbd): Block storage. Openshift will create an ext4 FS on top of it for you by default. Provide fast performance, but can only be accessed by one node at a time. - CephFS (storageClass: ocs-storagecluster-cephfs): Shared Filesystem storage. It's pretty much like NFS. Can be accessed by several pods simultaneously. - S3 (storageClass: openshift-storage.noobaa.io): Object storage. Require specific support from the application to use it. I don't think Redis support that kind of storage.
Basically, if you want more than 1 replica/pod to access your storage, pick CephFS. If you need max perf and single access (like a PSQL database), use RBD. If you want your data accessible from outside Openshift (i.e: on another VM), use NFS. With the exception of S3, all other storage types are used the same way. Only the PVC definition is different.
-darknao