I have 3 Kubernetes clusters (3 masters + 3 nodes per cluster), and I need to safely drain my Kubernetes nodes to upgrade the OS. I do this every week. All my nodes have a glusterfs daemonset : they are all storage node. But the last time, after I drained the nodes, heketi lost all the glusterfs cluster topology (On 2 of 3 Kubernetes clusters). The heketi database was empty. I don't know what happened, but I suspect a synchronization problem when restart heketi. The glusterfs cluster was OK.

So I have some questions : - Is there a way to resynchronize the heketi database from glusterfs cluster ? - I read I need to run "heketi-cli server mode set read-only" before restart heketi. Is there any other action I can do before drain a node (about storage) ?

Thank you.