Le 12/02/2019 à 18:36, Mildis a écrit :
Hi list,
haproxy is segfaulting multiple times these days for no apparent reason.
At first i thought is was a load issue but even few RPS made it crash.
Symptoms are always the same : segfault of a worker then spawn of a new.
If load is very high, spawned worker segfault immediatly.
In the messages log, the offset is always the same (+1e2000).
I'm running 1.9.4 (from vincent bernat package) in Debian stretch.
In haproxy logs :
Feb 12 11:36:54 ns3089939 haproxy[32688]: [ALERT] 042/113654 (32688) : Current
worker #1 (32689) exited with code 139 (Segmentation fault)
Feb 12 11:36:54 ns3089939 haproxy[32688]: [ALERT] 042/113654 (32688) :
exit-on-failure: killing every workers with SIGTERM
Feb 12 11:36:54 ns3089939 haproxy[32688]: [WARNING] 042/113654 (32688) : All
workers exited. Exiting... (139)
In /var/log/messages
kernel: traps: haproxy[32689] general protection ip:561e5b799375
sp:7ffe6fd3f2f0 error:0 in haproxy[561e5b72d000+1e2000]
"show errors" is empty.
How could I diagnose further without impacting production too much ?
Hi,
A recent fix about a double free has been merge in HAProxy 1.9:
http://git.haproxy.org/?p=haproxy-1.9.git;a=commit;h=451c5a88
Maybe you've hit this bug.
--
Christopher Faulet
