What is this alert?
This alert warns of high memory consumption on the VMware Greenplum segment hosts for an extended period.
Every 15 seconds, the VMware Greenplum metrics collector extension samples the memory in use on each segment host. Memory used for kernel cache and buffers is excluded. The average for all segments is calculated. If the average percentage remains above the threshold that is set for the alert for the number of minutes specified, an alert is issued.
If memory consumption is increasing over time, queries could start to faile with out of memory errors.
What to do?
Check the Query Monitor to see if there is unusually heavy query activity.
Look for active queries that perform hash joins or sorts on a large number of tuples. If possible, optimize the queries to eliminate rows earlier so that these memory-intensive operations process a fewer number of tuples.
Adjust resource queues or resource groups to limit the number of concurrent transactions.
Use the ps command to identify non-database processes consuming excessive memory. Stop unnecessary processes or move them to another server.
