virtual memory expanding

[emms]

We have a coldfusion application running on a server, using SQL, this is the only thing the server is used for.

We have recently had problems with memory, the virtual memory expands itself rapidly and uses up all the free space on the drive. the last time this happend we got down to 3MB and cfserver stopped responding. the only way we have found to stop this is by stopping and starting coldfusion, however this does not restore the memory used by the virtual memory.

are there any known problems like this? or can anyone suggest a way to stop this.

at the moment we have a 2GB drive (v.small but at the moment we're reluctant to put more in) and a 1.5GB paging file

I really am at a loss and would appreciate any help you can give me


thankyou

 

[Ghost]

emms,

Not sure if you have the same situation as us, but wanted to share what happened...

First, there are many reported memory leak issues with CF and some are documented on the

http://www.allaire.com

site. You should go there and read a little in their forums. The main issue is code that does not properly lock and unlock variables...

Our problem was the app we ran was fine on NT4, but wouldn't even run on Windows 2000 Server -- and still won't today. We opened an incident with Allaire tech support and went back and forth for 2 weeks. Bottom line is after EVERYTHING in our app was locked down for memory leaks when we flipped it to a W2K server is blew through ALL the memory within minutes (we have about 900 meg).

Allaire pointed the finger at Microsoft for W2K needing a patch or whatever. So, we kept our app on the NT4 server.

I'm VERY interested to see your feedback to this post.... If you are interested I'll get our lead programmer on this post to go over any details they may have.

Let me know.

Ghost

 

[emms]

Ghost,

thanks for your post, we've had a look at the locking and unlocking of variables and we're as sure as we can be that this isn't where the problem is.

The site has been running for several months on windows 2000 without a problem so I dont think this is the problem either.

After looking on the Allaire site we tried stopping and starting SQL Server on the machine, this freed up 84MB of disc space, so we are wondering whether this could be one of the problems.

After speaking to other cold fusion users I get the feeling that this is a common problem, and would appreciate any help from others who have experienced this - even if its just to tell me we're not alone in this!

Emma

 

[Ghost]

emms,

Just curious if you have had any further understanding of your issue. Curious too, do you have SQL and CF on the same machine?

We have it on different machines and do the stop/start always on the web server CF services to clear things up. We never have problems with SQL Server -- it's solid as a rock. We have an automatic restart every 24 hours but still get caught about once every two weeks having to do it for reasons we don't understand.

I'm also VERY surprised not one other CF shop has been motivated to post here. Maybe it's just us two???

Ghost

 

[Ghost]

I just heard from Allaire technical support shortly after release 5 comes out they will be releasing patches to 4.51 to address their memory leak issue as well as the issue why not all 4.51 sites run on Windows2000 Server. FYI.

Ghost

 

[TheCyniC]

You're not alone. I've had the same problem. We have 2 servers one for Oracle 8i and one for CFServer 4.5.1 and IIS. Last week we had to restart the web server when CFServer was using 111MB of memory, and had to restart it again when CFServer managed to tie up 432-stinkin'-MB of memory.

I've been coding Perl for 5 years now, and recently changed jobs. So far, I like Perl. If I find anything out, I'll post it, but it looks like our best hope is the patch that Ghost mentioned.

 

[Ghost]

theCynic,

Even if a recent change to your hardware or operating system has affected this problem I'd still go for an automatic stop and restart of services every 24 hours. Be sure to include a 10 second delay giving things a moment to shut down 100%.

Ghost

 

[rr236]

We've seen problems like this ever since we started using 4.51 rather than version 4.0

My understanding is that when 4.51 was introduced, Allaire decided to use a new version of smartheap, which doesn't release unused memory. The idea being that you'll eventually need it anyway.

We've also found problems with the enterprise version of 4.51, in that that version is not be able to efficiently manage native drivers. If you're using native drivers, and seeing problems with memory usage, I would recommend you use ODBC connections. Although ODBC leaks memory, CF seems to at least understand ODBC connections should be released when they aren't needed.

We changed over to ODBC, and although we have experienced a loss in performance, we at least now have servers that don't need to be rebooted every day.