Maintenance in Yandex StoreDoc

Maintenance includes:

• Automatic installation of DBMS updates and patches for hosts (including for stopped clusters).
• Other maintenance activities.

A major DBMS version update is not part of maintenance. For more information about version changes, see Yandex StoreDoc version upgrade.

Maintenance windowMaintenance window

You can set up a maintenance window when creating or updating a cluster:

• The arbitrary option (default) allows performing maintenance at any time.
• The by schedule option allows setting the preferred maintenance start day and time (UTC). For example, you can choose a time when the cluster is least loaded.

Maintenance workflowMaintenance workflow

The Yandex StoreDoc cluster maintenance workflow depends on the number of hosts and sharding.

Non-sharded clusterNon-sharded cluster

The maintenance procedure is as follows:

1. Secondary replicas undergo maintenance one by one. Such hosts are queued randomly. A secondary replica becomes unavailable while being restarted during maintenance.
2. After that, the primary replica (master) undergoes maintenance. If it is restarted and becomes unavailable, one of the secondary replicas will take its role. A single-host cluster will be unavailable during its maintenance.

Sharded clusterSharded cluster

In sharded clusters, the maintenance procedure is as follows:

1. The load balancer stops.

2. Hosts with the MONGOINFRA role for standard sharding or MONGOCFG for advanced sharding undergo maintenance one by one. For hosts with the MONGOINFRA role, only MongoCFG is subject to maintenance. Host maintenance is run in the same way as it would in a non-sharded cluster:

   1. Secondary replicas undergo maintenance one by one. Such hosts are queued randomly. A secondary replica becomes unavailable while being restarted during maintenance.
   2. Then, the primary replica undergoes maintenance. If it is restarted and becomes unavailable, one of the secondary replicas will take its role.

3. Shards undergo maintenance one by one, in ascending order by shard number. Host maintenance in each shard is the same as in non-sharded clusters:

   1. Secondary replicas undergo maintenance one by one. Such hosts are queued randomly. A secondary replica becomes unavailable while being restarted during maintenance.
   2. Then, the primary replica undergoes maintenance. If it is restarted and becomes unavailable, one of the secondary replicas will take its role. A single-host shard will be unavailable during its maintenance.

4. Hosts with the MONGOINFRA role for standard sharding or MONGOS for advanced sharding undergo maintenance one by one. For hosts with the MONGOINFRA role, MongoS is subject to maintenance. Such hosts are queued randomly. Any host that requires a restart during maintenance will be unavailable until maintenance is complete.

5. The load balancer resumes its operation.