ODP failure

OceanBase Database Proxy (ODP) failures are classified into the following two types: failures of an ODP node, and failures of the F5 load balancer, Linux Virtual Server (LVS), Alibaba Cloud Server Load Balancer (SLB), and other load balancing components at the upper layer of a cluster of multiple ODP nodes. This topic describes the emergency procedure to eliminate failures of an ODP node.

Emergency procedure

Failures of an ODP node may occur in the following two circumstances:

• The obproxy process frequently exits due to the out-of-memory (OOM) exception.

  If the obproxy process exits due to insufficient memory, you can find the following error message in the ODP log: obproxy's memory is out of limit's 90%. In this case, you can modify the proxy_mem_limited parameter, which specifies the maximum memory space that can be used by ODP. If the memory usage of ODP exceeds the specified value, the process exits. The value range is [100 MB, 100 GB]. The modification to the parameter immediately takes effect without a restart.

  If you deploy ODP by using OceanBase Cloud Platform (OCP), this parameter is set to 2 GB by default during the setting of the startup parameters. You can set the parameter to 8 GB.

  If you use a CLI tool to install ODP without explicitly specifying the memory size, the proxy_mem_limited parameter is set to 800 MB by default. You can modify the parameter to 2 GB.

  ALTER proxyconfig SET proxy_mem_limited = 2G;
  

• ODP cannot automatically restart due to exceptions of the daemon process.

  In this case, you can manually restart ODP.