SCIENTIFIC-LINUX-USERS Archives

November 2006

SCIENTIFIC-LINUX-USERS@LISTSERV.FNAL.GOV

Options: Use Monospaced Font
Show Text Part by Default
Show All Mail Headers

Message: [<< First] [< Prev] [Next >] [Last >>]
Topic: [<< First] [< Prev] [Next >] [Last >>]
Author: [<< First] [< Prev] [Next >] [Last >>]

Print Reply
Subject:
From:
"White, NGH (Nick)" <[log in to unmask]>
Reply To:
White, NGH (Nick)
Date:
Wed, 15 Nov 2006 16:59:03 +0000
Content-Type:
text/plain
Parts/Attachments:
text/plain (71 lines)
Hi,

The timeline for all these comments were within the first 3 months of 2006.
The situation may have changed since. We used a SL4.2 based system on an
Intel platform with an Areca card controlling 8 drives. Two being a mirror
for the OS and 6 being used as the data partition using RAID6. 

XFS likes 8K stacks. There were several threads around regarding XFS and 4/8K
stacks. 32 bit kernel used 4K stacks whilst 64 bit kernels used 8K stacks.
This is under review as kernel folks would like to use 4K stacks as that aids
performance.  

The basic problem is to do with stack overflow or just running out of stack
space. Recursive calls eat the stack space up. Calls through md and lvm to xfs
eat the space.

eg 
> > > > After building a couple of clusters using xfs on the shared storage 
> > > > device
> > > > (and using md and lvm on top of that), I'm getting this error now
which hard
> > > > crashes my machines:
> > > > 
> > > >  do_IRQ: stack overflow: 284
> > > >   [<c01078a2>] do_IRQ+0x44/0x130


Testing during the first 3 months of this year found the following:

1. To get 8K stacks in the 32 bit kernel is possible but you have to apply 
a patch to apic.c in order to get the kernel to compile.

2. Testing revealed that we were able to fill up a 1TB array but the AIMVII
benchmark failed within a minute with a "No resource available" kernel message.

3. Had to backport 3 fixes from the 2.6.11 kernel(current at the time) into SL4.2, 
which then enabled AIMVII to run successfully.

4. Then ran NFS tests into the test server, which threw up a stack issue regarding
low memory. Basically the stack overflowed as shown above. The server kept going 
but I would not run a production service on this configuration not only due to the 
stack issue but also having a high kernel maintenance overhead.

Regs.

- Nick 


>-----Original Message-----
>From: [log in to unmask]
>[mailto:[log in to unmask]]On Behalf Of Urs
>Beyerle
>Sent: 15 November 2006 10:13
>To: [log in to unmask]
>Subject: XFS on SL4
>
>
>Hi,
>
>As written in 
>ftp://ftp.scientificlinux.org/linux/scientific/44/i386/contrib/
RPMS/xfs/README , XFS 
is not recommended on SL4 32bit kernels.

What kind of problems where observed?
Is somebody working on a solution, which does not require to recompile the standard SL4 kernel?

Thanks,

Urs

ATOM RSS1 RSS2