Perl listener can fail to clean up children during post-crash restart

[Galen Charlton]

If a Perl listener throws an uncaught exception (e.g., like bug 1953044), OpenSRF::System->run_service() will try to clean up and reset the service. This cleanup includes terminating and reaping all children, then spawning new ones.

However, the reaping process doesn't quite work: OpenSRF::Server->cleanup() sets the CHLD signal handler to 'IGNORE', which means that Linux will automatically reap the child... meaning that ->reap_children() will not detect that they are gone and will leave them on the active and/or idle lists. When the service resets itself, it will think that it already has one or more drones ready to go, which in turn means that it may artificially hit max_children sooner than it needs to.

OpenSRF 3.1+

[Galen Charlton (gmc)]

A patch is available at the tip of

working/user/gmcharlt/lp1953057_forcibly_reap_after_reset / https://git.evergreen-ils.org/?p=working/OpenSRF.git;a=shortlog;h=refs/heads/user/gmcharlt/lp1953057_forcibly_reap_after_reset