Note
This parameter is available starting with V1.4.
Description
server_permanent_offline_time specifies the threshold for the time when a node is considered permanently offline. When a node is permanently offline, the Root Service processes the data replicas on that node, removes them from the Paxos group, and replicates the data to other available OBServer nodes to ensure the integrity of the Paxos group.
Privilege requirements
Query the parameter
The
systenant and all user tenants can query this parameter by using theSHOW PARAMETERSstatement or theGV$OB_PARAMETERSview.Modify the parameter
Only the
systenant can modify this parameter. User tenants cannot modify this parameter.
Attributes
| Attribute | Description |
|---|---|
| Type | Time |
| Default value | 3600s |
| Value range | [20s, +∞) |
| Modifiable | Yes. You can use the ALTER SYSTEM SET statement to modify the parameter. |
| Effective upon OBServer node restart | No. The setting takes effect immediately. |
Considerations
By setting this parameter, you can avoid unnecessary replica replication. The system only starts the replication of data replicas on permanently offline nodes when the replicas are lost for a period longer than the specified threshold.
The applicable scenarios and recommended values of this parameter are as follows:
Database version upgrade: We recommend that you set this parameter to
72h.Note
When you use OCP V2.3.5 or later to upgrade OceanBase Database V2.1.2 or later, the system automatically sets this parameter to
72h.Hardware replacement of an OBServer node: We recommend that you set this parameter to
4h.Clearing and online deployment of an OBServer node: We recommend that you set this parameter to
10mto enable the cluster to go online quickly.
Examples
Set the threshold for the time when a node is considered permanently offline to 3600s.
obclient> ALTER SYSTEM SET server_permanent_offline_time = '3600s';
References
For more information about the Root Service, see Database-level high availability.