|
|
The change in https://bugs.gentoo.org/show_bug.cgi?id=584410 only worked because of
multiple bugs interacting.
Due to variable shadowing https://bugs.gentoo.org/show_bug.cgi?id=639164 the loop did not
actually work, and the /run/haproxy.pid was used in every loop iteration.
This only worked because of a bug in openrc: https://bugs.gentoo.org/show_bug.cgi?id=639218
which was fixed in 0.34.11
So before it killed all processes called haproxy if nbproc > 1 in the config
(which means multiple haproxy instances are all killed if one is stopped!)
After the bugfix it doesn't kill, so we're back to the original situation of leftover
processes. And we don't want to rely on the cgroup killer!
The fix is obviously to use start-stop-daemon directly instead of calling openrc
internals that don't do what you think they do.
Package-Manager: Portage-2.3.16, Repoman-2.3.6
RepoMan-Options: --force
|
Patrick Lauer
GitWeb