Summary information |
|
|---|---|
Default value |
1 |
Valid values |
1 (enabled), 0 (disabled) |
Status |
Dynamic |
Display level |
Comprehensive |
Required role |
System administrator |
Configuration group |
Shared Disk Cluster |
Setting automatic cluster takeover to 1 allows an instance that is starting to automatically recover from an abrupt total cluster failure. If you set automatic cluster takeover to 0, the cluster may not be able to recover from an abrupt cluster failover unless you include the --cluster_takeover parameter.
The Cluster Edition uses quorum heartbeats and a cluster takeover algorithm to determine when cluster takeover should be performed. This algorithm allows an instance that is starting to distinguish between an inability to join the cluster because the cluster has crashed (in which case takeover is appropriate) and an inability to join the cluster because the instance that is starting does not have network connectivity (in which case takeover is not appropriate).
If you disable automatic cluster takeover (set it to 0), The Cluster Edition writes the results of the algorithm to the error log as an advisory message and then exits.
If you enable auotomatic cluster takeover (set it to 1), the Cluster Edition starts as the cluster coordinator and recovers the databases. This is guaranteed to be a safe operation in environments that have I/O fencing enabled.
In environments without I/O fencing, a malfunction of the algorithm could introduce data corruption, so you can set the configuration parameter to 0 to disable this algorithm. However, environments without I/O fencing have a risk of data corruption, and disabling automatic cluster takeover does not mitigate all of those risks.
