Skip to contentSkip to conversation

VM problems in 2.4.20

DISCUSSION Adam McKenna 2 participants 3 messages Feb 26, 2003

lore
Share & tools
lore
#0 lore raw

I'm having a VM issue on one of my servers running 2.4.20.

The box reports the following:

adam@foxy:~$ cat /proc/meminfo 
        total:    used:    free:  shared: buffers:  cached:
Mem:  4041986048 3968524288 73461760        0 619249664 1551077376
Swap: 2001215488 171663360 1829552128
MemTotal:      3947252 kB
MemFree:         71740 kB
MemShared:           0 kB
Buffers:        604736 kB
Cached:        1494600 kB
SwapCached:      20124 kB
Active:        1632528 kB
Inactive:      2006628 kB
HighTotal:     3080176 kB
HighFree:        67324 kB
LowTotal:       867076 kB
LowFree:          4416 kB
SwapTotal:     1954312 kB
SwapFree:      1786672 kB

As you can see there is plenty of memory sitting in buffers/cache. The problem is that when our daily cronjobs run, the box starts swapping and the load goes up to 30. The cronjobs are just the normal system cronjobs like updatedb, checksecurity, etc.

I had this problem a while ago with 2.4.6-xfs and 2.4.14-xfs, but this is stock 2.4.20 and I was under the impression that the VM was relatively OK by now.

adam@foxy:~$ grep -i mem /boot/config-2.4.20
# CONFIG_NOHIGHMEM is not set
# CONFIG_HIGHMEM4G is not set
CONFIG_HIGHMEM64G=y
CONFIG_HIGHMEM=y
# Memory Technology Devices (MTD)
# CONFIG_BLK_DEV_UMEM is not set
# CONFIG_DEBUG_HIGHMEM is not set

Any suggestions?

Thanks,

--Adam

#1 lore raw

On Wednesday 26 February 2003 20:40, Adam McKenna wrote:

Hi Adam,

Quoted 1 line

I'm having a VM issue on one of my servers running 2.4.20.

well, the vanilla VM, hmm, sorry: sucks :) for large boxen.

Quoted 8 lines

As you can see there is plenty of memory sitting in buffers/cache. The problem is that when our daily cronjobs run, the box starts swapping and the load goes up to 30. The cronjobs are just the normal system cronjobs like updatedb, checksecurity, etc.
I had this problem a while ago with 2.4.6-xfs and 2.4.14-xfs, but this is stock 2.4.20 and I was under the impression that the VM was relatively OK by now.

Quoted 9 lines
adam@foxy:~$ grep -i mem /boot/config-2.4.20
# CONFIG_NOHIGHMEM is not set
# CONFIG_HIGHMEM4G is not set
CONFIG_HIGHMEM64G=y
CONFIG_HIGHMEM=y
# Memory Technology Devices (MTD)
# CONFIG_BLK_DEV_UMEM is not set
# CONFIG_DEBUG_HIGHMEM is not set
Any suggestions?

yes, use either -rmap (1) or -aa (2).

1.) www.surriel.com/patches (Rik van Riel)
2.) www.kernel.org/pub/linux/kernel/people/andrea/kernels (Andrea Arcangeli)

There is also a recent thread you may want to read:
- http://marc.theaimsgroup.com/?l=linux-kernel&m=104617534110371&w=2

ciao, Marc

#2 lore raw
Quoted 6 lines · Marc-Christian Petersen

On Wednesday 26 February 2003 20:40, Adam McKenna wrote:

Hi Adam,

Quoted 1 line

I'm having a VM issue on one of my servers running 2.4.20.

well, the vanilla VM, hmm, sorry: sucks :) for large boxen.

Can you expand on "sucks"? 4GB is not really that large of a system anymore.

I still don't see how this memory taken up by buffers/cache is not freed when a system process needs it. I have always thought that this type of memory is by definition "lower priority", and should be dropped when a process needs more memory. Is this not the case?

--Adam