GPDB 6.x: Enabling gp_enable_gpperfmon causes the database to go into recovery mode
search cancel

GPDB 6.x: Enabling gp_enable_gpperfmon causes the database to go into recovery mode

book

Article ID: 296757

calendar_today

Updated On:

Products

VMware Tanzu Greenplum

Issue/Introduction

When using GPDB 6.x with GPCC 6.x, it is not required to turn on gp_enable_gpperfmon GUC because GPCC no longer uses the gpmmon and gpsmon agents to collect metrics information. GPCC 6.x internally has it's own metrics collector.

If gp_enable_gpperfmon is enabled in GPDB 6.x, it can lead to the following error:
2022-03-15 14:25:46.509647 EDT,,,p300647,th-813553536,,,,0,,,seg-1,,,,,"WARNING","01000","ic-proxy-server: received signal 3",,,,,,,0,,"ic_proxy_main.c",379,
2022-03-15 14:25:46.442932 EDT,,,p300525,th-813553536,,,,0,,,seg-1,,,,,"LOG","00000","3rd party error log:
2022-03-15 14:25:46|:-FATAL: [INTERNAL ERROR gpmmon.c:1080] agg_dup failed
2022-03-15 14:25:46.484928 EDT,,,p300524,th-813553536,,,,0,,,seg-1,,,,,"LOG","00000","worker process: perfmon process (PID 300648) exited with exit code 1",,,,,,,0,,"postmaster.c",3952,
2022-03-15 14:25:46.484948 EDT,,,p300524,th-813553536,,,,0,,,seg-1,,,,,"LOG","00000","terminating any other active server processes",,,,,,,0,,"postmaster.c",3691,
gpmmon logs will report a similar message. This causes the database to crash and go into recovery mode which would lead to a short outage.

Environment

Product Version: 6.18

Resolution

gp_enable_gpperfmon is deprecated and VMware currently recommends that customers do not turn it on as it can lead to errors like this.
To check the current setting:
gpconfig -s gp_enable_gpperfmon

To disable the GUC, run the following command and restart the database:
gpconfig -c gp_enable_gpperfmon -v off