The Lotus Cars Community banner

1 - 14 of 14 Posts

·
Registered
Joined
·
2,564 Posts
Discussion Starter #1
My file server is leaking about 8K every 5 seconds. I have my suspisions as to what is causing the leak, but I can't confirm it. If I top the machine, I see the used and/or active memory increase 8K every 5 seconds, but if I sort my processes by memory usage, I don't see any processes' memory footprint increase.

suggestions?

FYI: the server is running RedHat 9 with the 2.4.20-8smp kernel (it *has* to run that kernel as that is the only kernel my promise RAID card works under. :mad: bah).
 

·
Registered
Joined
·
2,195 Posts
BrianK said:
My file server is leaking about 8K every 5 seconds. I have my suspisions as to what is causing the leak, but I can't confirm it. If I top the machine, I see the used and/or active memory increase 8K every 5 seconds, but if I sort my processes by memory usage, I don't see any processes' memory footprint increase.

suggestions?
Do you suspect a user-space task or a kernel driver leak? You can keep an eye on /proc/slabinfo and /proc/meminfo to watch what the kernel is up to.

vmstat will show overall memory usage over time.

For tasks, you can use
'ps aux'
and watch what they have to resident (RSS) and virtual size (VSZ)


FYI: the server is running RedHat 9 with the 2.4.20-8smp kernel (it *has* to run that kernel as that is the only kernel my promise RAID card works under. :mad: bah).
Well that there is the problem with proprietary raid controllers. Either get a well-supported one with driver source (like 3ware?) or dump the card, connect your drives directly, and let Linux manage the RAID for you.
 

·
Registered
Joined
·
2,564 Posts
Discussion Starter #3
Re: Re: Anyone know how to find the source of a memory leak in Linux?

Ground Loop said:
Do you suspect a user-space task or a kernel driver leak? You can keep an eye on /proc/slabinfo and /proc/meminfo to watch what the kernel is up to.
vmstat will show overall memory usage over time.
hmm... slabinfo shows no change. memifo shows a change in buffers, active, activeanon, activecache. vmstat shows buffers increasing.

What does the meminfo report tell me? There's not a whole lot of documentation on it out there.
For tasks, you can use
'ps aux'
and watch what they have to resident (RSS) and virtual size (VSZ)
yeah, that shows no change in anything.

Well that there is the problem with proprietary raid controllers. Either get a well-supported one with driver source (like 3ware?) or dump the card, connect your drives directly, and let Linux manage the RAID for you.
agreed, unfortunately, they don't advertize the fact that it only works under that one old kernel. It was half the price of the 3ware (my original choice) - now I see why. I'd rather not go with software raid as this thing gets hammered on pretty hard.
 

·
Registered
Joined
·
2,195 Posts
Looks like you're not the only one having memory allocation problems. EliseTalk has been puking these up all evening:

Fatal error: Allowed memory size of 134217728 bytes exhausted (tried to allocate 184320 bytes) in /usr/local/psa/home/vhosts/elisetalk.com/httpdocs/forums/showthread.php(489) : eval()'d code on line 84

Fatal error: Allowed memory size of 134217728 bytes exhausted (tried to allocate 6 bytes) in /usr/local/psa/home/vhosts/elisetalk.com/httpdocs/forums/admin/functions.php on line 2407
 

·
Forum Founder
Joined
·
29,081 Posts
Yeah, I rebooted the server. I have been dealing with a memory leak also. Very similar.
 

·
Registered
Joined
·
6,802 Posts
Man Randy, up a bit early today!

Also, regarding the huge traffic spike yesterday. You don't think it was a hacker type thing do you?
Chris
 

·
Registered
Joined
·
4,985 Posts
I think you do it like you would with a suspect innertube. Mix up a bucket of very soapy water and dip your RAM in it. Keep an eye out for a stream of fine bubbles. There's your leak!
 

·
Registered
Joined
·
609 Posts
Should be obvious

Pull up your task manager and look for a process called "Elise_soft_top" - that'll be leaking for sure...:)
 

·
insert clever title here
Joined
·
7,702 Posts
Tough problem. I'd suggest killing processes one at a time until you see the leaked memory restored.
 

·
Registered
Joined
·
368 Posts
Might I suggest converting to Windows Server 2003. I have no leaks in any of my Windows 2003 servers. In the corner I have an NT4 box that has been running virtually non-stop since 1997...

:)
 

·
Registered
Joined
·
687 Posts
Debugging things like memory leaks in programs with at least tens of thousands of lines of code is a big reason why I chose not to be a programmer--too much stress, and even if the work day is over, you are still thinking about it...

Anyway, if the file server is important to you, get mainstream supported parts so that you can run the latest and greatest.
 

·
insert clever title here
Joined
·
7,702 Posts
Later Apex said:
In the corner I have an NT4 box that has been running virtually non-stop since 1997...
If true, that must be a world record for a Windows machine! ;) Seriously though, how is that possible? Windows is notoriously insecure, needing frequent patching and updates. These patches and updates often/usually require a reboot. Are you saying you haven't installed any updates, nor installed any new software, since '97?
 

·
Registered
Joined
·
368 Posts
MattG said:
If true, that must be a world record for a Windows machine! ;) Seriously though, how is that possible? Windows is notoriously insecure, needing frequent patching and updates. These patches and updates often/usually require a reboot. Are you saying you haven't installed any updates, nor installed any new software, since '97?
Nobody attacks a circa 1997 machine... :) that or dumb luck...


As for finding memory leaks in NT, you can use debug mode to do it and attach a debugger and it will pretty quickly show you what driver or application is the culprit. We nailed IBM to the wall on a faulty PCI driver for one of their network cards once using that procedure. They furnished new drivers ASAP. :)
 
1 - 14 of 14 Posts
Top