Unable to start up GPTEXT after reboot

Products

• VMware Tanzu Greenplum

Issue/Introduction

Due to a known issue in gptext version 3.11.0, when starting or stopping gptext the Oprtating System's OOM-killer (Out Of Memory killer) may kill the gptext processes.

See Managing Out of Memory states for details on the oom-killer.

Resolution

Workaround

To bring up the gptext cluster:

1. List the nodes in the GPText cluster
   SELECT * FROM gptext.index_status();
   
   
2. Start one node using the -n option:
   gptext-start -n <node>
   
   
3. Run gptext-recover to start the rest of the cluster
   
   
4. Run gptext-state to check the current state
   It may take some time for the recovery to finish.
   Use the psql to connect to the database and run "SELECT * FROM gptext.index_status();" to monitor the recovery status.

Fix

Upgrade to GPText v3.11.1 or above to get the fix for the issue.

See Release Notes for details:

Greenplum Text no longer reloads an index after a utility activates or deactivates the index readonly state. This change helps to avoid OOM errors caused by suddenly reloading indexes when the Solr node JVM memory usage is over 50%.