> [...]
> Anyway, I have a pretty vanilla ubuntu host (openvpn virtual host).
> I installed pcp on it and have been noticing the memory use being high.
> Pretty small host, 2GB memory but 47% feels a bit much for pmmgr.
> PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
> 1270 pcp 20 0 1349932 979308 2440 S 0.0 47.7 3400:22 pmmgr
> [...]
The RES column sounds like a problem. Please describe the pmmgr
version/configuration, lifespan of the process, pmmgr.log file,
and whatever else looks relevant.
The TIME+ column looks way too high also, as though there was a
tight failure/retry loop in effect sometime. (commit 90ea027d3
fixes one possible trigger for this.)
(The VIRT column is not necessarily a problem. The confluence of
pmmgr & libpcp probe=NNN discovery multithreading has been seen to
fragment the virtual address space pretty easily, but the vast
majority of that space is unmapped. It may be possible to ease
this particular aspect, but it's a separate, non-leak issue.)
A more typical pmmgr top(1) report looks like this:
PID USER NI VIRT RES SWAP vMj %CPU nTH TIME+ COMMAND
16477 pcp 0 4224352 6856 916 0.5 1 2:00.46
/usr/libexec/pcp/bin/pmmgr -l pmmgr.log
or this:
PID USER NI VIRT RES %CPU TIME+ SWAP nFLT COMMAND
3188 pcp 0 99220 2400 0.0 1:02.01 640 63
/usr/libexec/pcp/bin/pmmgr -l pmmgr.log
- FChE
|