Hi, On 03/06/2024 09:39, Remi Gacogne wrote:
On 02/06/2024 22:43, Nicolas Baumgarten via dnsdist wrote:we started testing 1.9.4 on centos 7 before upgrading, and it segfaults, after 20 seconds of running.Thanks a lot for reporting this issue! I'll try to reproduce this on CentOS 7 later today, as it doesn't happen in my development setup or in our CI. In the meantime, would you by any chance be able to get a backtrace ("bt full" in gdb) from a core dump, if any has been generated? Or, if you still have the setup laying around and can spare a few minutes, by starting dnsdist inside gdb: "gdb dnsdist" then "run", and once you get to the crash "bt full".
In case anyone else stumbles upon this, the issue seems to be caused by a bug in the net-snmp version shipped by EL7. I can reproduce it by opening a very large number of file descriptors before starting the SNMP thread, which causes the net-snmp provided "run_alarms" function to segfault because it uses a deprecated net-snmp function that does not support large file descriptors (hence the warning in the logs). I tried to reproduce this behaviour with a more recent version of net-snmp (5.9.4) and I couldn't, so for now I don't think I'll bother spending more time on it. If you are running into this issue, please check how many TCP worker threads you are starting with setMaxTCPClientThreads, as it really doesn't make sense to start more than a dozen of them in most cases since 1.4.0 and legacy configurations sometimes use a very high number there.
Best regards, -- Remi Gacogne PowerDNS.COM BV - https://www.powerdns.com/
OpenPGP_signature.asc
Description: OpenPGP digital signature
_______________________________________________ dnsdist mailing list dnsdist@mailman.powerdns.com https://mailman.powerdns.com/mailman/listinfo/dnsdist
