0

I´m having a strange memory trouble with a Linux box with two JBoss instances and one Tomcat instance. All was working fine until last 20/jul. After that, the memory available (total used - buffers - cached) has decreased from 3-4GB to under 500MB. I added 6 GB RAM to this server but the result was the same.

This is the memory configuration of each instance actualy, with 16 GB RAM server:

Tomcat (7.0.22): Heap - 4G, MaxPermSize - 1G, Jboss inst1 (5.1.0.GA): Heap - 1 GB, MaxPermSize = 512 Mb Jboss inst2 (6.0.0): Heap - 3 GB, MaxPermSize = 512 Mb

No new versions of any app was deployed on server.

Distribution: Debian 5.0.9 Kernel: 2.6.26-2-amd64

If I kill those process, the SO releases only about 6 GB of RAM, as we can see:

apora:~# free -m; kill -9 17431; sleep 10; free -m
             total       used       free     shared    buffers     cached
Mem:         16090      16004         86          0          0         13
-/+ buffers/cache:      15989        100
Swap:         4580       2301       2279
             total       used       free     shared    buffers     cached
Mem:         16090      12597       3492          0          0         13
-/+ buffers/cache:      12583       3506
Swap:         4580       1439       3141
apora:~#

Killing Jboss Inst02:

apora:~# kill -9 20385
apora:~# free -m
             total       used       free     shared    buffers     cached
Mem:         16090      10908       5182          0          1         21
-/+ buffers/cache:      10885       5204
Swap:         4580        723       3857
apora:~#

Killing Jboss Inst01:

apora:~# free -m
             total       used       free     shared    buffers     cached
Mem:         16090      10207       5883          0          1         21
-/+ buffers/cache:      10183       5906
Swap:         4580         33       4547
apora:~#

I killed other processes like nscd, dsmc (tivoli TSM agent), and others that was consuming memory according to top, but my system still consuming about 10 GB of RAM.

Result of cat /proc/meminfo shows a little of Slab used and reclaimable:

apora:~# cat /proc/meminfo
MemTotal:     16476656 kB
MemFree:         87588 kB
Buffers:        233468 kB
Cached:         265064 kB
SwapCached:          0 kB
Active:        5876332 kB
Inactive:       304664 kB
SwapTotal:     4690928 kB
SwapFree:      4690304 kB
Dirty:             516 kB
Writeback:           0 kB
AnonPages:     5682444 kB
Mapped:          50244 kB
Slab:           152660 kB
SReclaimable:   118892 kB
SUnreclaim:      33768 kB
PageTables:      26768 kB
NFS_Unstable:        0 kB
Bounce:              0 kB
WritebackTmp:        0 kB
CommitLimit:  12929256 kB
Committed_AS: 10292968 kB
VmallocTotal: 34359738367 kB
VmallocUsed:    171948 kB
VmallocChunk: 34359565951 kB
HugePages_Total:     0
HugePages_Free:      0
HugePages_Rsvd:      0
HugePages_Surp:      0
Hugepagesize:     2048 kB

Does anyone know what is happening ?

Thanks,

André

André Sousa
  • 1
  • Although the system and kernel are pretty old, was the kernel version changed around the time this started happening? There seem to have been some memory leaks in the kernel along the way, not that I'd know of such in the particular one you mentioned. – zagrimsan Aug 18 '15 at 07:10
  • A suggestion related to the above: would it be possible to try to upgrade the kernel? If the kernel version you quoted is the linux-image package name, there aren't any more recent kernel binaries available for Debian 5, but if you have a spare environment, you could first copy everything into it, verify that the issue still is reproducible, and then compile e.g. the last available 2.6 kernel from sources and see if that makes any difference. – zagrimsan Aug 18 '15 at 07:21

0 Answers0