Accounting replication issue
aland at deployingradius.com
Thu Jul 27 17:02:38 CEST 2017
On Jul 27, 2017, at 10:15 AM, Chris Adams <linux at cmadams.net> wrote:
> I have two servers running FreeRADIUS (CentOS 7 with freeradius-3.0.4).
> They are logging accounting records to local MySQL servers. For
> redundancy, I configured FreeRADIUS to replicate the accounting records
> to each other, based on the robust-proxy-accounting sample config.
> This works, except when one server is unavailable (call it server A),
> the other (server B) stops replicating the records until I restart
> FreeRADIUS on server B. Server B logs that server A went away, and logs
> that it came back (in this case, somebody did something dumb on server A
> and triggered the kernel oom_killer - didn't touch the FreeRADIUS or
> MySQL processes, just the dumb grep):
> Jul 24 09:25:38 srvB radiusd: Marking home server srvA port 1813 as zombie (it has not responded in 6.000000 seconds).
> Jul 24 09:25:38 srvB radiusd: (52212975) ERROR: Failing proxied request, due to lack of any response from home server srvA port 1813
> Jul 24 09:25:42 srvB radiusd: No response to status check 52213042 for home server srvA port 1813
> Jul 24 09:26:05 srvB radiusd: No outstanding request was found for reply from host srvA port 1813 - ID 197
> Jul 24 09:26:06 srvB radiusd: No outstanding request was found for reply from host srvA port 1813 - ID 198
> Jul 24 09:26:06 srvB radiusd: (52213350) Received response to status check 52213350 (1 in current sequence)
> Jul 24 09:26:06 srvB radiusd: (52213350) Marking home server srvA port 1813 alive
That's all good...
> The comments in the sample config just say "When the home servers come
> back up, the packets are forwarded, and the detail file processed as
> normal." - I don't see any configuration related to that, but it
> defintely did not happen for me; the records were just written to the
> detail file until I restarted the service (which then processed the
> backlog correctly).
> Any suggestions for what to look at?
Upgrade to 3.0.15. It's been found and fixed a long time ago.
More information about the Freeradius-Users