I am trying to understand Session Management feature of WVE.Requesting your assistance.
In WVE, APC makes runtime decisions regarding which cluster members to be started/stopped .So here applications can lose their sessions even without any explicit application server failure .This frequency would be fairly high when compared to that in WAS ND topology.
• Would like to know how efficient session management is in WVE
• How to identify the overhead associated with it
• Does a substantial portion of available system resource get spent on session management in WVE.
• Is "HTTP session rebalancing" feature of WVE supported, only when you are using ObjectGrid-based HTTP session support , and not incase of the normal WebSphere(R) Application Server session management.
KeithS 060000MSV417 Posts
Re: Session Management in WVE2012-02-27T14:57:34ZThis is the accepted answer. This is the accepted answer.Paresh,
VE doesn't do it's own session management except for session re-balancing, but it is aware of how many sessions are active when it stops a server. This applies to both when it stops a server due to a health policy action as well as when it stops a server in a dynamic cluster for performance reasons. In both cases, it will prevent new sessions from being established on that appserver and wait up to some maximum amount of time for existing sessions to expire before stopping the appserver. If they expire before this interval, then the appserver will be stopped sooner.
So in response to your 3 questions:
1) The additional overhead is minimal. We require PMI statistics be enabled to get the active session count, and for session rebalancing, the DWLM controller (one per cluster) calculates when re-balancing should occur. This is very small amount of overhead.
2) No, the overhead is small.
3) Yes, session rebalancing is supported even w/o ObjectGrid/eXtreme Scale.
SystemAdmin 110000D4XK102 Posts
Re: Session Management in WVE2012-02-27T17:46:26ZThis is the accepted answer. This is the accepted answer.
- KeithS 060000MSV4
Thank you for ur reply .It cleared up a lot of things for me.
Please help me better understand one point that you have mentioned.
It being that, if any member servers have to be stopped due to health policy action,performance issues etc, it waits for a certain amt of time for existing sessions to expire before stopping the app server.
How does this work.Does it waits for a predetermined interval for all sessions to terminate,else it stops the server and recreates the killed sessions on another member server.
Also, is it proper to say that ,as this frequency of stopping/starting of servers in WVE is more when compared to WAS ND, so the session management in WVE plays a more vital role (along with increased resource consumption ) when compared to that in WAS ND.