What were you trying to do that didn't work?
pmlogger_daily seems to take more than one full day while executing pmlogcheck on the PCP data. This makes 1 full CPU be used constantly.
The directory of the data is 4GB big, trying to execute pmlogger_rewrite with the customer data, I can confirm at least more than 2 hours were required, I stopped after 2 hours of execution.
Please provide the package NVR for which bug is seen:
pcp-5.3.7-17.el8_8.x86_64
How reproducible:
Always on customer system
Additional informations
The customer is actually running RHEL8.4 latest PCP pcp-5.2.5-6.el8_4.
With this release, pmlogger_daily executions start accumulating because of the way the pmlogger_daily.service is implemented (Type=oneshot + KillMode=none + TimeoutStartSec=1h), which causes numerous pmlogcheck processes to consume up to all the CPUs.
Latest PCP for RHEL8.8. changed pmlogger_daily.service implementation, which now avoids accumulating processes, still it's not expected that pmlogcheck execution takes more than 1 hour, as suggested by the TimeoutStartSec=1h property in the service unit, which BTW does nothing at all since service has Type=exec).
- links to
-
RHBA-2023:124474 pcp bug fix and enhancement update