This topic describes possible exceptions related to hardware and network other than node failures, network jitter, and OceanBase Database Proxy (ODP) faults, and provides corresponding emergency procedures.
Emergency procedure
In addition to node failures, network jitter, and ODP faults, hardware exceptions include disk I/O exceptions, disk recognition failures, network interface controller (NIC) exceptions, and CPU or memory exceptions at nodes. These hardware exceptions often lead to an increase in the response time (RT) and failure of transaction commit.
To confirm a suspected hardware exception, run the stop server command. The stop server operation is lossless and safe and is the most effective method for fault isolation. However, an error is returned if a client accesses the isolated node. To prevent the error, run the isolate server command before the stop server operation, if possible. Examples:
ALTER SYSTEM ISOLATE SERVER 'xxx.xxx.xxx.xxx:2882';
ALTER SYSTEM STOP SERVER 'xxx.xxx.xxx.xxx:2882';
If you confirm that the database errors are caused by a hardware exception, replace the faulty server. For more information, see Replace an OBServer node.