Leader switchover exceptions

When the election leader detects an exception on the local machine, it triggers an abnormal leader switch. The following scenarios are included:

• CLOG_DISK_HANG: The clog disk hangs for more than 5s. You can adjust the tolerance time by setting the cluster parameter log_storage_warning_tolerance_time.
• SLOG_WRITE_HANG: The slog-writing thread hangs for more than 5s. You can adjust the tolerance time by setting the cluster parameter data_storage_warning_tolerance_time.
• SSTABLE_IO_HANG: The data disk hangs for more than 5s. You can adjust the tolerance time by setting the cluster parameter data_storage_warning_tolerance_time.
• CLOG_DISK_FULL: The clog disk is full.
• SCHEMA_NOT_REFRESHED: The schema is not refreshed.

When the election leader detects any of the preceding exceptions on the local machine, it lowers its election priority and selects a replica with a higher priority to trigger a re-election. You can query the DBA_OB_SERVER_EVENT_HISTORY view to check abnormal leader switch events and the corresponding causes.

obclient> select * from DBA_OB_SERVER_EVENT_HISTORY where module='FAILURE_DETECTOR' limit 10;

You can query the DBA_OB_SERVER_EVENT_HISTORY view to check the corresponding election events.

obclient> select * from DBA_OB_SERVER_EVENT_HISTORY where module like '%ELECTION%' limit 10;