-
Bug
-
Resolution: Done
-
High
-
El Alto Release
Hello,
today when deploying Modeling with OOM, the PersistentVolumeClaim needs the "ReadWriteMany" (or "RWX") capability:
$ kubectl get pvc -n onap | grep modeling onap-modeling-modeling-genericparser Bound onap-modeling-modeling-genericparser 2Gi RWX 3h25m onap-modeling-modeling-mariadb Pending 3h25m onap-modeling-modeling-mariadb-data-onap-modeling-modeling-mariadb-0 Bound onap-modeling-modeling-mariadb-data1 2Gi RWX onap-modeling-modeling-mariadb-data 3h25m onap-modeling-modeling-mariadb-data-onap-modeling-modeling-mariadb-1 Bound onap-modeling-modeling-mariadb-data0 2Gi RWX onap-modeling-modeling-mariadb-data 3h23m onap-modeling-modeling-mariadb-data-onap-modeling-modeling-mariadb-2 Bound onap-modeling-modeling-mariadb-data2 2Gi RWX onap-modeling-modeling-mariadb-data 3h22m
According to Kubernetes Documentation (https://kubernetes.io/docs/concepts/storage/persistent-volumes/#access-modes), ReadWriteMany stands for "the volume can be mounted as read-write by many nodes".
That means that a particular PVC needs to be read and written from many pods. That also means that your code takes that into account and do the work to avoid write at the same place at the same time.
An issue on RWX mode is that most "official" storage driver from Kubernetes doesn't support it (13 over the 19 drivers doesn't support it, espacially OpenStack, Amazon and Google storage classes).
It's then very hard in a lot of case to use Modeling.
If the needs is that the volume can be mounted as read by many nodes and write by one node (which seems to be the case), RWO (ReadWriteOnce) should be sufficient and is supported by all drivers.
in the effort to deploy ONAP for gating on public cloud (one the priority of El Alto), using storage classes (and if possible RWO access) is very important.
Can you verify that RWO is sufficient and if yes, I can submit a Gerrit review to move your access needs to RWO.