Segfault on FR 3.0.4 when MySQL server goes away

[Alan DeKok]

On Nov 24, 2015, at 7:19 AM, Jonathan Gazeley <Jonathan.Gazeley at bristol.ac.uk> wrote:
> 
> I'm currently running a mixed estate of RADIUS 2 and 3 on CentOS 6 and 7 (halfway through a migration project). The older servers are running a home-built FR 2.2.9 and the newer ones are running stock FR 3.0.4, as supplied with CentOS 7.
> 
> I've noticed the FR 3 boxes segfault from time to time and I think I've pinned it down to odd behaviour when the MySQL server is temporarily unavailable. We have clustered SQL behind an F5 load balancer but there is a small delay during failover when one of the SQL nodes goes away or when the F5 decides to switch to a different node for some other reason. The FR2 boxes cope fine with this transition but the FR3 boxes segfault in about 50% of cases.
> 
> I understand 3.0.4 is not the latest and I'm happy to build 3.1.x if this will fix the problem but initially I'm trying to work out if this is a bug in the server, or a misconfiguration on my part (although I'd hope a misconfiguration wouldn't cause a crash).
> 
> Backtrace included below, although this was not built with --enable-developer since it is a busy production system. I tried to replicate this on a dev system but the much lower load makes it hard to cause the crash.

  I'd use 3.0.10.  It has all known bugs fixed.

  Alan DeKok.