OceanBase Database is a distributed database in which the call links are complex. When a timeout issue occurs, O&M engineers cannot quickly locate whether the issue is caused by OceanBase Database components or the network. They can only analyze the issue based on experience and OBServer logs. The end-to-end diagnostics mechanism is provided in this version to help users quickly locate internal issues of database components.
There are two end-to-end paths. In one path, the application sends a request to OceanBase Database Proxy (ODP) by using a client, such as JDBC or OCI, to access the OBServer. The access result is returned to the application. In the other path, the application directly accesses the OBServer by using a client, and the access result is returned to the application. In end-to-end diagnostics, all components on the link are diagnosed. You can access OceanBase Database by using a client, and the client request is forwarded to an OBServer node by ODP. Alternatively, you can directly access the OBServer node. O&M engineers can use related methods in the PL/SQL package DBMS_MONITOR to enable applications to determine whether to enable end-to-end diagnostics trace based on different identifiers, and whether to record the trace information in trace logs. Trace logs are printed based on the access path. Related logs are recorded in the corresponding log files of ODP and OBServer for access by using ODP. For direct access to an OBServer node, logs are recorded only in the corresponding OBServer log file. The trace log file of ODP is named obproxy_trace.log, and that of OBServer is named trace.log. Each trace log file is 256 MB in size. When the size of a trace.log file reaches 256 MB, this file is archived and a new log file is generated. The number of archived log files to be retained is specified by the archive parameter. When the maximum allowed number of archived files is exceeded, earlier files in excess of the limit are deleted. O&M engineers can collect and analyze all trace logs to trace information such as the execution time of each transaction or SQL statement in the entire access link, thereby locating issues in the entire link.