All of lore.kernel.org
 help / color / mirror / Atom feed
* Relieving lowmem pressure on a highmem box, 2.4
@ 2003-06-25 19:43 Ian Soboroff
  2003-06-25 20:09 ` Martin J. Bligh
  0 siblings, 1 reply; 2+ messages in thread
From: Ian Soboroff @ 2003-06-25 19:43 UTC (permalink / raw)
  To: linux-kernel


I have a Dell server with 12GB of RAM in it which is having a lot of
trouble with lowmem pressure.  wli's bloatmeter script shows that it's
a lot of buffer-heads and inodes.

$ bloatmost | head -4
         buffer_head:   191316KB   211893KB   90.28
         inode_cache:    64813KB    66097KB   98.5
             size-64:       82KB    10871KB    0.75
             size-32:       95KB     5694KB    1.68

$ cat /proc/slabinfo | sort -k 2,2n | tail
size-512            1052   1176    512  132  147    1 :  124   62
dentry_cache        1068   2580    128   86   86    1 :  252  126
blkdev_requests     1200   1200    128   40   40    1 :  252  126
filp                1230   1230    128   41   41    1 :  252  126
size-32             1532  91118     64   42 1571    1 :  252  126
pte_chain           2466  13050    128  157  435    1 :  252  126
size-128            2946   3450    128  115  115    1 :  252  126
vm_area_struct      3192   4830    128  138  161    1 :  252  126
inode_cache       129503 132195    512 18885 18885    1 :  124   62
buffer_head       2043732 2260200     96 56450 56505    1 :  252  126

What's a good solution for this?  I'm not ready to move to 2.5, since
stability is pretty important for us, but patching 2.4 should be OK.
The current kernel is RedHat's 2.4.20-18.8 (SMP, BIGMEM; HIGHMEM
option set to 64GB).

Thanks in advance,
Ian


^ permalink raw reply	[flat|nested] 2+ messages in thread

* Re: Relieving lowmem pressure on a highmem box, 2.4
  2003-06-25 19:43 Relieving lowmem pressure on a highmem box, 2.4 Ian Soboroff
@ 2003-06-25 20:09 ` Martin J. Bligh
  0 siblings, 0 replies; 2+ messages in thread
From: Martin J. Bligh @ 2003-06-25 20:09 UTC (permalink / raw)
  To: Ian Soboroff, linux-kernel

> I have a Dell server with 12GB of RAM in it which is having a lot of
> trouble with lowmem pressure.  wli's bloatmeter script shows that it's
> a lot of buffer-heads and inodes.
> 
> $ bloatmost | head -4
>          buffer_head:   191316KB   211893KB   90.28
>          inode_cache:    64813KB    66097KB   98.5
>              size-64:       82KB    10871KB    0.75
>              size-32:       95KB     5694KB    1.68
> 
> $ cat /proc/slabinfo | sort -k 2,2n | tail
> size-512            1052   1176    512  132  147    1 :  124   62
> dentry_cache        1068   2580    128   86   86    1 :  252  126
> blkdev_requests     1200   1200    128   40   40    1 :  252  126
> filp                1230   1230    128   41   41    1 :  252  126
> size-32             1532  91118     64   42 1571    1 :  252  126
> pte_chain           2466  13050    128  157  435    1 :  252  126
> size-128            2946   3450    128  115  115    1 :  252  126
> vm_area_struct      3192   4830    128  138  161    1 :  252  126
> inode_cache       129503 132195    512 18885 18885    1 :  124   62
> buffer_head       2043732 2260200     96 56450 56505    1 :  252  126
> 
> What's a good solution for this?  I'm not ready to move to 2.5, since
> stability is pretty important for us, but patching 2.4 should be OK.
> The current kernel is RedHat's 2.4.20-18.8 (SMP, BIGMEM; HIGHMEM
> option set to 64GB).

Try 2.4-aa kernel - it copes much better with buffer_head bloat in
most circumstances. 

As long as your drivers are supported, 2.5 is actually mind-bogglingly 
stable on the whole - more so for large ia32 machines than any 2.4 
kernel I've seen (exactly because of lowmem issues).

M.


^ permalink raw reply	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2003-06-25 19:55 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2003-06-25 19:43 Relieving lowmem pressure on a highmem box, 2.4 Ian Soboroff
2003-06-25 20:09 ` Martin J. Bligh

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.