OpenVZ Forum

Home » General » Support » Kernel cache dentry leak? (Kernel CentOS 6 2.6.32-042stab044.11 and 2.6.32-042stab044.17)
Kernel cache dentry leak? [message #44943] Sun, 15 January 2012 17:56 Go to previous message
Messages: 11
Registered: January 2012
Junior Member

After setup node server with Centos 6 and 64bit latest openvz kernel 2.6.32-042stab044.11 we have noticed, that even on empty node after the two days node memory usage is shown 40-50%! After the inspection of running processes we have not noticed any process which uses significant amount of memory.
After the searching google we found that there may be kernel cache problem, so we executed command:
sync && echo 2 >/proc/sys/vm/drop_caches
And all used memory is freed.
The problem is, that these in cache used memory is reported as "used" memory, not a reusable cache or etc?
Does someone exeperience this kind of problem? If yes, maybe there is any solution?
Or maybe this is an incorrect memory usage reporting problem, or this is memory leak?
We have no such problem in all our centos 5 32bit kernel nodes. On Centos 5 32bit nodes there is no incressing memory usage (by cache or etc).

Also, on other forum we have an answer that someone else got this problem too:
We are experiencing the same thing.
Linux 2.6.32-042stab044.11 #1 SMP Wed Dec 14 16:02:00 MSK 2011 x86_64 x86_64 x86_64 GNU/Linux

Current RAM usage:
free -m
             total       used       free     shared    buffers     cached
Mem:         23971      13559      10411          0        330      12444
-/+ buffers/cache:        784      23187
Swap:        26622          6      26616

After sync:
free -m
             total       used       free     shared    buffers     cached
Mem:         23971       1068      22902          0        330        458
-/+ buffers/cache:        279      23691
Swap:        26622          6      26616

There is currently 1 VM running on this node. Why in the world would the server use 12 GB RAM as cache?

slabtop command shows big and incressing "dentry" usage.
I see by running "slabtop", that "dentry" value is very big and incressing continually. After cache clear it drops, but then incresses and incresses again and again.
For example, on our nominal loaded Centos 5 32 bit node there is around 700k objects in "dentry" and uses around 106MB.

In this time, on the empty node with Centos 6 64 bit kernel "dentry" holds 7000k objects (x10 more than on our loaded node!) and uses 1.5GB RAM and incresses...

Thank you for any help and answers.

[Updated on: Fri, 20 January 2012 21:36]

Report message to a moderator

Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Previous Topic: vzctl RHEL6 Kernel - No ploop module (Squeeze)
Next Topic: IO Problems
Goto Forum:

Current Time: Thu Dec 07 06:47:21 GMT 2023

Total time taken to generate the page: 0.02000 seconds