Living in Fear of JRun (w/CFMX)

I can't say that we've found a fix yet but we've been fighting a similar problem for about 2 months. Our environment is different (CFMX 6.1 for J2EE, Solaris, Apache) but the problem is the same... memory spikes and it all dies.

We've updated every conceivable peice of software, tweaked settings,  modified tons of code and still we have problems. Initially we were crashing multiple times per day... at the moment, we can make it about a week before it dies. Macromedia is going through all our logs every day and I think they've run out of ideas. At this point we're basically taking guesses. The graphing engine is suspect at the moment.

I'll try to put together a list of the things we've done so far, perhaps one of them will help you.

Everynow and then our ram will spike and Jrun takes over all CPU usage on our test server, which runs dual xp athlons IIRC. We still have no solution but I think I have traced it back to something between coldfusion and our SQL server. Sometimes it will even throw and error 4096. I've search for the error but have come up with nothing...

we've found that CFMX doesn't handle loss of connectivity to a database very well. it seems the threads never get released... requests queue... memory goes bye-bye... server dies.   it's pretty reliable that if our database goes down, the CF server is soon to follow.  We primarily use Oracle db's but do have a few MS SQL & DB2 databases.

from what ive seen - JRUN does take about 47% of ur memory ... no matter what machines u have !

so even if u increase ur RAM - it wld still hover around 40 - 50% mark !

Same problem, we've gone up to 3GB of RAM in our dual 2Ghz Xeon server, performance is STILL an issue!  Any ideas please let me know!

I know one thing we did on our servers that helped with startup issues and some all around performance issues is we had our simutanious connections set to some where up in the 200-300 range which is a bit high...ok a lot high. We set it to 40 and things seem to go a lot smoother on restarts. In addition if you are using MX, which I assume we all are, if you use a lot of c++ cfx that will slow performance since JRUN and the JVM have to go through more to use the CFX, converting those to Java instead will increase performace. Same thing goes for COM objects coldfusion just runs slow when you use COM objects. I read some articles on it a while back if I can find them I'll put them up here.

To me, in a lot of ways CF is better than ASP but no one else can consider this platform with this issue.  It's not even a minor, uncommon issue.  It's very common.  The Macromedia boards are plagued with this very same issue.  Macromedia doesn't really help at all.  This is the context of most posts regarding this issue:

- User: "Memory leak, Jrun taking up 600 MB, etc"
- Macromedia: "Post your settings, configurations and process dumps"
- User: Posts all requested information.
- User (2 months later): "Hello?"

That's the general response from Macromedia.  It's very frustrating.

Which version of JVM are you using?

Here's a fix for some SQL server problems.
http://www.macromedia.com/support/coldfusion/ts/documents/cfmx61_sqlserver_cpu.htm

any progress on this?

My main issue is with the cpu getting pegged at 90 somehting percent every so often, I have been talking with someone at Macromedia off and on for  over 3 weeks, trying to get my issue resolved with memory leaks, and cpu overutilization. (going over configs, sending in stack traces, etc) to no avail, we still have a major issue with jrun in CFMX and our current workaround is to auto restart cfmx every 4-6 hours, which is no good!

we've managed to rememdy the problem over here but I couldn't honestly tell you what "fixed" it. In the end, our sys-admins said the server was overloaded. We moved some of our sites to different servers to distribute the load a bit and the problem went away. The reason I'm not entirely sure this was the problem is that we also made several other changes at the same time.

Is it possible to find out what other changes you made?
I have isolated one site on a new clean server, with a clean database server as well and still see the problem, guessing a magic setting change may help.....

I'll see what I can find out from everybody. we had no less than 5 people working on this so it was pretty much a big mess.

FYI Jrun seems to have settled down on our servers now, having applied every updater and patch and driver under the sun.  We even got it running pretty fast on a knackered old PIII 600 server, and we've been able to scale down our live servers from 3GB of RAM to 2GB.  All our servers are running Windows 2K Advanced Server with SP4, SQL 2K with SP4, MX on Jrun with all the latest patches.  

No acceptable answer provided. Refund and close.

i'd recommend PAQ/refund

The Jrun issue on our servers was code related and was fixed. I have confirmed with others that have not posted that this was their problem as well. Try putting the ball back in the developers court. The stack traces actually do have vauable info in them, get with macromedia support and have them point out the bad cfc or lines of code from the traces...... Good Luck.  CFMX is actually not to bad when it's running clean code.

-Brian

i agree there is no clear answer (or any "answer" for that matter) but i think it worth having the conversation archived. I don't see any harm in leaving it.

Are you instantiating any COM objects?  If so, these will cause pagefile leaks, which, in my experience, will cripple a server.

If you've got 6.1, use the ReleaseComObject() function on every COM object ref as soon as you're finished with it.

Fixed my problem, which I had for way too long, so I figured I'd share... ;-)

-Liam

We are having the same problem I have a massive server, running Windows 2003, IIS6. MX with 10GB or RAM.  Everytime the jrun.exe gets over 512mb it just stops responding.  I've checked the exceptions log and see all these out of memory errors.  I have to restart the server completely, then watch the memory keep going up and up.

Any help would be great, I hate restarting this server everyday.

-Chuck

Been watching this post for some time...  And hopeing to find an answer...  what I have come up with (Since this is also a problem on our servers as well), is that there is no ONE thing that is causing this...  Macromedia says it's just problems with how you code your pages...  but there doesn't seem to be a way to actually pinpoint where the memory leak is actually occuring...

I see errors in the application log that says just "Occured in template...." without even an explination of WHAT the error might have been...

These appearing around the time that the memory leak decides to trash JRUN...

Incidentally, don't think you have to run your memory up to 512 or in our case as high as 1.4meg before JRUN will crash...  I've seen JRUN get trashed and the memory just hovering around 200K...  this is a little rediculous and getting to be a huge problem...

turning off debugging on our production server seems to have fixed our problem!