Hide Forgot
Created attachment 1216710 [details] Patch to spec file Description of problem: While rebuilding memcached for RDO, we found out this issue in upstream CI. Upstream ticket has been created: https://github.com/memcached/memcached/issues/218 Version-Release number of selected component (if applicable): memcached-1.4.33-2.fc26 How reproducible: Always Steps to Reproduce: 1. https://gist.github.com/dmsimard/174971182870d03394f3f2223cf59412 Alternatively 1. change limits in /etc/sysconfig/memcached 2. systemctl restart memcached Actual results: crashes Expected results: shoyuld run Additional info: We fixed it in RDO package, distgit patch is attached
Note that in these systemd times, it would probably be best to use "User=memcached" from the unit file, as with the increased LimitNOFILE the process doesn't need to be started as root to be able to increase that system value itself. The upside is that the CapabilityBoundingSet change is no longer relevant. The downside is for people editing the sysconfig file to run memcached as a user different from "memcached"... I honestly think no one does that, and even then, it would be as easy to add a User=foo than it is to modify the sysconfig file. I have been using User=memcached myself for a while, because I also add RuntimeDirectory=memcached in order to use UNIX sockets created inside /run/memcached/ instead of TCP. And I had needed to increase LimitNOFILE too.
This bug appears to have been reported against 'rawhide' during the Fedora 26 development cycle. Changing version to '26'.
It seems this was addressed in 1.4.34 by adding the CAP_SYS_RESOURCE capatibility. Thanks!