We are using Webfocus version 7.7.01 on our test machine because we are intending to move production to 7.7.01. One of the problems that we have is that when we exit our maintains the agents do not release the process. The state shows in use and never releases. The agent has to be killed. The workspace is setup like this:

SERVICE = DEFAULT
BEGIN
maximum = 40
number_ready = 10
agent_refresh = 1
END

The way we exit maintains is this:

function OnbtnExit_Click ( ) {
IWCTrigger("WINEXIT");
window.close();
}

Any ideas?

This message has been edited. Last edited by: Kerry,

Could you please let me know what browser / version you are using?

Thank you!
Mark

Internet Explorer 8

Ok. I have been informed that there is a fix for this issue and that you should open a case and request it. The fix should stop the agents from staying up.

Mark

Wha? I'm on 7.7.03 and still seeing this behavior. Is there a fix that hasn't propagate to subsequent versions?

Hi John
We are testing here in 7703m, which was the maintenance release for 7703, and we are not seeing this issue. Are you on that release or can you upgrade to it?

Is there is a easy was for us to repro your issue? You could send me the files and I can make sure that it will not be a problem in 7703m.

Please let me know how you want to proceed.

Thanks
Mark Derwin
Mark_Derwin@ibi.com

We're on 7703. There seem to be certain screens that have the problem more than others, and I'm looking into seeing which IDs are associated with the hung screens to see if that can give me some insight into the behavior causing the trouble.

This is a big Maintain install, so it's not a simple send-you-a-file kind of thing. I was hoping you'd say set Connection_Limit to 3000 or something like that.

This hasn't been an issue in the past because we've simply bounced the servers at night and given the 60 max agents for the day. The machine runs like a top even with 50+ agents available. But the VM machine it's on is having memory issues and is a bit ornery when it restarts, so they've turned off a weekly reboot and for some unknown reason turned off the bouncing of the services. So we've run out of agents a couple of times and I'm revisiting the problem.

If there's not a setting involved I'm going to see if I can figure out what user behavior is resulting in the most hung jobs. They read as "In Use" but can linger for days without disappearing.

Someone at IB also recommended having each agent only answer one request prior to being dropped years back, which may exacerbate any memory leak issue in Reporting Server or Win2k8. We go through a few thousand agents in a typical day. Just not set up as generally recommended by IB.

J.

John
Do you use session_idle_limit? That might help clear the hanging agents if users forget to close the application.

What platform are you running on? Also, if you could send us the edaprint.log and edasprof.prf we will be happy to take a look.

Mark

We do have a session_idle_limit set but they're not idle -- they're listed as in-use and don't appear affected by the idle limit. I'm running a couple of experiments that take a few days to play out, but I may take you up on the edaprint and edasprof offer.

Win 2k8 R2 by the way.

Stunning update -- in 7.7.03M changing idle_session_limit will indeed end agents that are in the "in use" state. I've set it to 36000, 10 hours, and I now have a clean slate each morning. I no longer have to worry about running out of agents due to "in use" agents holding open slots.

This more or less solves my problem, though the server I'm on appears to have a memory leak. No indication it's Reporting Server yet, though the settings on this machine have resulted in a few thousand new agents per day. I'm putting those settings back to more normal states to see if there's a change on Win 2k8.

J.

i know this is an old thread, but i'm thinking about setting my idle_session_limit.
But what is 'idle'?
If an agent is running a big oracle extract.... and the user is not interacting with it..just waiting... is that agent 'idle'??
If I set the idle_session_limit, would it kill an agent that is still cranking data???

John, when you say 'will indeed end agents that are in an in use state', would it kill someone's active job if its a long one?