Search <book_title>...

NetBackup™ Deployment Guide for Kubernetes Clusters

Last Published: 2025-02-26

Product(s): NetBackup & Alta Data Protection (10.5.0.1)

Decoupling of NetBackup Policy and Job Management from primary server

In Cloud Scale Technology, policy execution and job management (NBPEM and NBJM) is implemented as StatefulSet which can be deployed through the operators. Policy and job management capabilities are decoupled from the primary server and run as separate service in Kubernetes cluster.

See Deploying the operators.

NBPEM/NBJM is dependent on primary, web services and mqbroker pods. NBPEM/NBJM cluster comprises of one leader pod and one or more follower pods. StatefulSet policyjobmgr represents the leader whereas StatefulSet policyjob is associated with the followers.

Figure: Policy and Job Management architecture

Note:

In the above figure, NB Web services is a decoupled component. For more information refer to Decoupling of NetBackup web services from primary server.

A NBPEM/NBJM pod consists of following containers:

The main container for core policy and job management
PBX sidecar
Vnetd sidecar
Fluentbit sidecar
Housekeeping sidecar

nbpemreq

The nbpemreq (NetBackup Policy Execution Manager Requisition) command shows the distribution of policy/clients across the follower pods.

For Cloud Scale, it is recommended to run the nbpemreq command on the primary pod or on the policyjob leader.

For more information on nbpemreq command, refer to the NetBackup™ Commands Reference Guide.

Limitations and recommendations

Multiplexing (MPX): MPX is not supported in NBPEM/NBJM scale-out environment.
Supporting MPX in any format could lead to unpredictable/unexpected/undesired behaviors. Hence it is recommended to disable the MPX support.
Session end and catalog backup: For scale-out deployments (Kubernetes cluster), session end catalog backup schedules are not supported.
For non scale-out deployments, session end catalog backup schedules and persistence would be moved to a Json formatted file.
Scale-out/Scale-in:
- Auto-scaling is not supported. However, manual scaling of policyjob StatefulSet is allowed. The policyjobmgr StatefulSet cannot be scaled.
- By default, Cloud Scale starts with a single replica of policyjob follower pod; using the scale-out operation, this can be increased to a maximum of 127 replicas.
  Warning:
  The users are advised to scale out the policyjob follower service only if necessary as this would consume more resources.
- Run the following command to control the replica count:
  kubectl -n <namespace> patch environment nbux --type merge --patch '{"spec": {"primary": {"replicas":{"policyjob":'<COUNT>'}}}}'
  A sample script to control the replica count is as follows:
```
# IMPORTANT: Initialise NAMESPACE to the namespace in which Cloud Scale is deployed
 
$ COUNT=$1
$ PJSTS=$(kubectl -n $NAMESPACE get sts | egrep '\<policyjob\>' | awk '{print $1}')
$ kubectl -n $NAMESPACE get sts $PJSTS &> /dev/null || { echo "StatefulSet $PJSTS not found..."; exit 1; }
$ (( replicas = $(kubectl -n $NAMESPACE get sts $PJSTS -o jsonpath='{.spec.replicas}') ))
$ (( replicas == $COUNT )) && { echo 'Requested replica count same as current...'; exit 0; }
$ kubectl -n $NAMESPACE patch environment nbux --type merge --patch '{"spec": {"primary": {"replicas": {"policyjob":'$COUNT'}}}}'
```