Outstanding requests increasing until max_outstanding

[Louis Munro]

Hello,

I have a pair of FreeRADIUS (2.2.5) servers acting as load-balancers for a pool of authentication servers behind them that has been running flawlessly for about a year except for one thing.

When I look at the list for my auth servers using radmin I get the following: 

radmin> show home_server list 
10.0.0.100	1812	auth	alive	150
10.0.0.101	1812	auth	alive	223
10.0.0.102	1812	auth	alive	130
10.0.0.103	1812	auth	alive	114

Where I believe the last column is the number of outstanding requests for each home server (please enlighten me if I am wrong about that).

Now the problem I have is that that last number (of outstanding requests) never goes down significantly.
It fluctuates a bit during the day, but the trend is clearly up. 

This morning it was showing the following...
radmin> show home_server list 
10.0.0.100	1812	auth	alive	140
10.0.0.101	1812	auth	alive	220
10.0.0.102	1812	auth	alive	109
10.0.0.103	1812	auth	alive	114

And yesterday it was 


radmin> show home_server list 
10.0.0.100	1812	auth	alive	102
10.0.0.101	1812	auth	alive	30
10.0.0.102	1812	auth	alive	78
10.0.0.103	1812	auth	alive	75

And last Friday one of the counters hit 65535 and all proxying to that home server stopped even though it was still responding (and was marked “alive”).

I believe that counter is incremented once every time a request is proxied and (should be) decremented every time a reply is received or it times out.
It does look though like there may be a case where the counter is not decremented.

So my questions are: 

1. Do I understand this correctly? 
2. Is this a known issue? 
3. If so, is it fixed in the latest of v2? or v3? 
4. If it’s not an issue, what could I be doing wrong? 

I have no problem with the answer being “Upgrade” as long as I know that this is something that is indeed fixed in the version I’ll be upgrading to.
These servers have performed so well up to now that I have not had a compelling case for upgrading. This could be just it.

Regards, 
--
Louis Munro
lmunro at inverse.ca  ::  www.inverse.ca 
+1.514.447.4918 x125  :: +1 (866) 353-6153 x125
Inverse inc. :: Leaders behind SOGo (www.sogo.nu) and PacketFence (www.packetfence.org)