I try to run my report and it uses up all the EDA agents. Why would my report run on multiple agents?

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

if you are in an early version of 71
(i think it was 3 but i can't remember now), there was a problem with a certain type of execution opening up a zillion agents, eating up the machine. it was pretty funny. i don't remember the specifics, i had opened a case, and the situation was replicated, and that version was quickly replaced with the next service pack. so, your post doesn't say what version/platform you're running. Please update your automatic signature in your profile.

When you cancel your browser session before a report is run it does not automatically cancel the agent.

Possibly this common misunderstanding might explain why this situation is occurring.

Regards

John

quote:

eating up the machine

We had a misterious case that the memory usage jumpted from normal 88% to 99% and almost crashed our system. We stopped the Webfocus server and the memory usage went back to normal. That probly happended 3-4 times sporadically. We didn't pay attension to how many agents were active at the time.

We also experienced what John described. Does a crashed agent consider an active agent?

Hua

Does anybody know if there is a way to add an Abort button to the launch page that will kill or stop the agent being used. We have several users that get impatient if a report takes too long, so they will close the browser window and restart the report, causing the first agent to hang in use, and starting a second agent for the new request. This then causes performance issues for the rest of the user base.

Thanks

We've seen a report eat up all agents a few times when one of our database servers wasn't responding¹. It accepted connections and queries just fine, it just didn't return results in any timely fashion and eventually locked up all available agents waiting for their results from that database.

That wasn't a single issue of the report being run though, which appears to be what the OP is seeing?

quote:

Originally posted by Hua:
We also experienced what John described. Does a crashed agent consider an active agent?

A crashed agent is a different story really, but I can confirm that we've seen reports trying to use a crashed agent instead of an available one, which can lead to some unnecessary and rather fruitless debugging if you're not aware of that.

quote:

Originally posted by MO Admin:
Does anybody know if there is a way to add an Abort button to the launch page that will kill or stop the agent being used.

That won't help you, those users aren't going to press an Abort button either.
It is possible though that their new request for the same report ends up in the same (webserver-)session. Perhaps you could somehow figure out the agent ID that handles your report, store that ID and an identifier for the report being run (for example, &FOCFEXNAME) in your session (the foccache should work, I think) and if you find that same report identifier in your session when starting the report, you kill the related agent (-SYSTEM kill &PID) if it is still executing your report. That's a pretty long list of mights and maybes, I don't know if it's really possible (or viable).



1). It's an RDB installation on OpenVMS. WebFOCUS' queries on that which involve joins are so ridiculously slow that we usually just fetch entire table contents and join them in FOCUS - it's faster.

I believe I can get the users to use an Abort button if it were available before closing the browser or restarting the report. However, from what I have found (or I guess haven't found), it doesn't look like that is a possibility. Has anyone ever done this before?