CLUSTER FAILOVER
Since 1.1.0Cluster
Syntax
CLUSTER FAILOVER [FORCE | TAKEOVER]Sent to a replica: makes it the master of its shard, and its master a replica of it.
| Option | Effect |
|---|---|
| none | The master pauses writes, the replica catches up, and then it is elected. Nothing acknowledged is lost. |
FORCE | Skips the catching up, for a master that cannot be reached. An election is still held. |
TAKEOVER | Skips the election too: the replica takes the slots at once, without the other masters' agreement. For when no majority of masters can be reached. |
A master that fails is replaced the same way without this command: once a majority of masters have not heard from it for cluster-node-timeout milliseconds, its replicas stand for election. When it comes back, it finds its slots taken at a higher epoch and becomes a replica.
Reply
OK, before the failover is done.ERR You should send CLUSTER FAILOVER to a replica.
Examples
> CLUSTER FAILOVER
OK
> ROLE
1) "master"
...