|
Hi Matt,
I'm wishing to run Jrun on my realservers. The only thing is that Jrun
is an application server, which will intriduce sessions based up on
connnections, and the sessions will get stored in memory. If I have a
bunch of realservers, how do they all know about the sessions and what's
in memory? I might have the concept a little skewed, but I'm prettty
sure this is how it works. If anybody can explain to me how this works,
I'd appreciate it greatly.
I run a topology using Websphere which store session info into memory. When
you run this kind of topology, you need to enable persistence connection on
your LVS director. Another alternative is to dedicate a centrale database
who store the session informations directly instead of using websphere
memory. This solution introduce a SPOF on the session database but session
handling sollicitation is minor in front of other critical data. This is a
typical 3/3 architecture...
=> Solution is use persistence if you only need to be sure a specific
client connection still stuck to the same realserver, you need to enable
jrun session timeout lower than LVS realserver persistence timeout.
=> Another solution is Using a centralized database dedicated for jrun
application session handling => So persistence on LVS is not needed.
Best regards,
Alexandre
|
