New Comment on Issue #github

[pcp:master] New Comment on Issue #118 reported pmDestroyContext: pthread_mutex_destroy(c_lock) failed: Device or resource busy, failed second try as well ( https://github.com/performancecopilot/pcp/issues/118 )
By kmcdonell ( https://github.com/kmcdonell ) :

Frank, thanks for the detective work.

This may be related to the relatively recent "disconnect" logic changes in libpcp that were aimed at cases where pduread() being interrupted, but potentially implicate any error patch from a socket read failure.

In your failure case (if it is reproducible), what's the callstack above __libc_recv() ... I'm guessing it leads to pduread() but I'm most interested in who's calling pduread() ... specifically is it a pmFetch() or a metadata operation or the handshake during pmNewContext()?

Also the pmmgr.log tail has entries from two days, I'm guessing only the Oct 20 ones are relevant to this problem.

And finally, does pmmgr create multiple PCP contexts for the same PMCD host (that may help me construct a simpler failure scenario outside pmmgr for QA).