Comment 44 for bug 1850281

Fernando,

the logs you sent me indicate pmlogger is starting but exiting very soon after (and then being repeatedly restarted by systemd). I don't see any evidence of a systemd unit configuration problem. The PCP archives are being created, but have a corrupted label record and I can't read any of them. e.g. here's what I get with some debugging enabled trying to examine one of your recent faulty archives :

$ pminfo -Dlog,logmeta -a 20200504.11.16-00
__pmLogOpen: inspect file "./20200504.11.16-00.0"
__pmLogOpen: inspect file "./20200504.11.16-00.index"
__pmFopen("./20200504.11.16-00.index", "r"): no decompress
__pmLogOpen: inspect file "./20200504.11.16-00.meta"
__pmFopen("./20200504.11.16-00.meta", "r"): no decompress
__pmFopen("./20200504.11.16-00.0", "r"): no decompress
__pmLogChkLabel: fd=3 vol=0 trailer read -> 4 (expect 4) or bad trailer len=8650752 (expected 132)
pminfo: Cannot open archive "20200504.11.16-00": Illegal label record at start of a PCP archive log file

This may indicate a bug in pmlogger / libpcp. Can you please try running pmlogger manually, e.g.
$ pmlogger -c config.default -h local: -l- -s 5 -t 1 /tmp/testlog

Please send me the output of the above command (-l- will print messages to stdout) and also /var/lib/pcp/config/pmlogger/config.default and also /tmp/testlog.*

If the above command fails, please: rm /var/lib/pcp/config/pmlogger/config.default and then systemctl restart pmlogger (this will regenerate a new config.default since yours may be corrupted somehow). After restarting the pmlogger service, please also send me the output of: systemctl status pmlogger; journalctl -xe -u pmlogger.service

Also, back in Comment #20, you said you have systemd.i686 241-12.git323cdf4.fc30 installed. Are you running a 32bit machine?

Regards
--