Maintenance in Managed Service for Redis
- Automatic installation of DBMS updates and revisions for hosts (including disabled clusters).
- Changes to the host class and storage size.
- Other maintenance activities.
Changing a major DBMS version isn't part of maintenance. For more information about major version changes, see Redis version upgrade.
- Unspecified time (default): Maintenance is possible at any time.
- By schedule: Set the preferred maintenance start time: desired day of the week and UTC hour. For example, you can choose a time when cluster load is lightest.
Maintenance procedure for Managed Service for Redis clusters depends on the number of hosts and presence of shards:
In non-sharded single-host clusters, maintenance is performed on a master host. So, such a cluster becomes unavailable if a master host needs to be restarted during maintenance.
If a non-sharded cluster is comprised of a few hosts, the maintenance procedure is as follows:
- Replica hosts undergo maintenance consecutively. The hosts are queued randomly. A replica becomes unavailable while it's being restarted during maintenance.
- Maintenance is performed on a master host. If it's restarted during maintenance and becomes unavailable, one of the replicas assumes its role.
In sharded clusters, maintenance is performed shard by shard in ascending order by shard number. Host maintenance is the same as in non-sharded clusters.