Cluster Maintenance- Services Downtime

[#1]

Hello ,

I administer a small production cluster (2NN, 7DN).

I want to perform a cluster maintenance (OS & Security updates) which will require for all nodes to be rebooted.

From what i read impala catalog and statestore services can be migrated to different nodes with no issues.

What about HiveServer, metastore and our databases (mariadb) (single host)?

Can their hosts be rebooted without overal cluster downtime or should we enable HA for them?

Ideally we want to avoid downtime.

[#2]

hello @pphot 

You can migrate HMS and HS2 instances to any other hosts. So you can add another hosts for HS2 and HMS instance and remove the previous ones once the new ones are added and are functioning normally.

For backend database, if you migrate then you have to update the configuration in the CM so that HMS has the updated information and access as to which host it needs to communicate in order to access its backend DB. Please note this is mandatory as Hive has all the information stored in its backend DB.

Let me know if the above helps.

Regards,

Tushar