Command Center checks the status of the Greenplum Database segments every 30 seconds and raises an alert if any segments are down or running in their non-preferred roles. This alert will be raised hourly until an administrator has recovered the failed segments.
With segment mirroring enabled, Greenplum Database can tolerate a primary or mirror segment failure as long as there is a working instance for every segment in the cluster. If both the primary and mirror instances fail for any single segment, Greenplum Database cannot run queries. For this reason, it is important to recover the failed segment instance to protect from loss of service.
Segment instances have a "preferred role," either primary or mirror. When all segment instances are in their preferred roles, each segment host has the same number of primary and mirror segments. If a primary instance fails, its mirror instance assumes the primary role and the distribution of primary segments is no longer balanced. This can slow down query processing because a host with more primary segments than others can take longer to complete queries.
Restore the failed segments and return the segments to their preferred roles so that the cluster is in balance.
See Recovering From Segment Failures for steps to recover Greenplum Database segments.
