We have a lvs configuration consisting of two machines, where each machine
acts as both master/backup director and realserver. The LVS handles http and
https requests. Everyhing works great except for some (very few) clients
which experiences a great latency. I have established that the latency only
occurs when connecting through the LVS (i.e. connecting to the VIP). If the
client connects directly to the real server that handled the request, no
latency can be seen. We have measured the time spent in the realserver, and
there everything is OK for both cases.
Does anyone have a clue of what may cause the latency? Is there any easy way
of measuring where the latency occurs?
Best Regards
Per Wicklund
|