Hi
I have about 50 web servers running RedHat 6.2 being served by a ServerIron.
Some of the real servers get "stuck" connections that never go away. These
accummulate over time on the load balancer and cause the traffic
distribution to be uneven. Sometimes the number of sessions will even go
down to zero yet the connections remain. The only way I have found to clear
them is to do a reboot. I have tried both connection balancing and session
balancing but neither one works properly.
The web server are identical in hardware, software and configuration. Yet
this only occurs to certain servers, not all. I also have another ServerIron
with the same kernel handling 100 servers that does not have this problem.
The only difference I can see between them is that this one has peak traffic
of around 85mbs whereas the other only gets up to around 40. Above 70 mbs
the difference between inbound and outbound start to seperate like there is
some sort of packet loss occuring. I am planning on upgrading it to a
gigabit connection but I am not sure if this will handle it.
My questions are:
Has anyone seen this type of behavior or have any ideas on what is causing
it?
Is there a way to view or clear the real server connections without
rebooting?
Thanks,
Bill McCaffrey
bill@neopets.com
This archive was generated by hypermail 2b29 : Mon Aug 04 2003 - 04:10:05 EDT