we assumed that the 'priv' and 'pub' RESTEnvironment types always contained an AnyEvent eventloop, but this is actually not the case in pvestatd and pvescheduler.
When we wrongly determined that, it depended on the used model that AnyEvent used (and autodetected) if it worked or not. With AnyEvent::Impl::Perl it did not make problems (and seemingly worked by accident), but when using AnyEvent::Impl::EV (which is autodetected and used when libev-perl is installed) it interfered with our SIG_CHLD handlers and only ever called them once. (Not clear why this happens, maybe because AnyEvent is not setup correctly). This patch uses $AnyEvent::MODEL as a detection instead since this is `undef` until the first AnyEvent watcher is created, which should be only the case where we really use AnyEvent, such as pveproxy and pvedaemon. Fixes: 6870afa ("RESTEnvironment: better SIGCHLD handling in AnyEvent event loop") Signed-off-by: Dominik Csapak <d.csa...@proxmox.com> --- src/PVE/RESTEnvironment.pm | 6 ++---- 1 file changed, 2 insertions(+), 4 deletions(-) diff --git a/src/PVE/RESTEnvironment.pm b/src/PVE/RESTEnvironment.pm index c258b1e..89def38 100644 --- a/src/PVE/RESTEnvironment.pm +++ b/src/PVE/RESTEnvironment.pm @@ -112,12 +112,10 @@ sub init { die "unknown environment type" if !$type || $type !~ m/^(cli|pub|priv|ha)$/; - my $has_anyevent = $type eq 'pub' || $type eq 'priv'; - $SIG{CHLD} = sub { - # when we're in an api server, we have to postpone the call to worker_reaper, otherwise it + # when we're using AnyEvent, we have to postpone the call to worker_reaper, otherwise it # might interfere with running api calls - if ($has_anyevent) { + if (defined($AnyEvent::MODEL)) { AnyEvent::postpone { $worker_reaper->() }; } else { $worker_reaper->(); -- 2.30.2 _______________________________________________ pve-devel mailing list pve-devel@lists.proxmox.com https://lists.proxmox.com/cgi-bin/mailman/listinfo/pve-devel