We have a number of available agents and when we first start them, the memory allocation is about 7004K per tscom3. When we run a report, the tscom3 jumps to 34567K. Even after the report is done and the agent is idle, the memory allocation does not go back to pre-execution size (7004K). What's worst is that we have a tscom3 with a size of 1053467K and no report is running.

Is there a way to automatically reset the allocation?

Thanks,
Dan

In workspace configuration - Data Services and the set of Agents you use there is idle limits etc etc to mess around with.

Be aware that the whole point of agents is to provide a swift response by having the resources on standby. You will have to make compromises (as always in this life!)

hammo1j,

Thanks for the reply.

Having 50 or even 100 agents on standy is not really a problem for us. The problem is when an agent is running and the memory allocated to that agent builds up to 104MB and not release it after it becomes idle again.

If for example we start with 50 agents and we set the memory limit to 150MB and it goes over, the workforce manager will kill that agent. We will be then have 1 less agent in stand-by. The Workspace Manager will not attempt to maintain the number of agents (from number_ready) after startup if agents are terminated for any reason.

We don't really want to terminate the agent. We want it to de-allocate the memory when it goes back to the idle state.

might just try shortening the idle agent limit. Small overhead because it may have to start a new agent, but any resources the agent controls (such as memory) are released when the agent is shut down after x seconds of "idleness" - a compromise as earlier suggested with a minimal overhead/delay if the agent stays idle for too long

Use the agent refresh limit option. That way you can assure that the agent is refreshed and also standing by so that it does not have to started when the request comes.

this would do basically the same thing. with one correction - agent refresh doesn't start a new agent. it just stops the old one after x number of connections. so is isn't standing by - it also has to start a new agent when requested

Thanks for the replies.

Darin,
If for example you got like 3 agents active and all of them are used most of the time and the idle limit value (say 1 hour) does not get triggerred, the agent memory allocation continues to grow. The manager will not get a chance to kill it and restart one (with a much lower memory allocation), unless of course you manually terminate it.

They ALL eventually get recycled due to agent refresh, but i think the default value is 100. Shorten that limit (along with idle limit) and you get something more useful

Darin,

That's exactly what I did in our QA environment. I cut it in half and notice the memory hovers at a lower level than before.

Thanks.