[Nottingham] LVM and load balancing and ReiserFS

[Duncan John Fyfe]

On Sun, 2007-10-28 at 17:56 +0000, Peter Chang wrote:
> On Sat, 27 Oct 2007, Duncan John Fyfe wrote:
> [...]
> >
> > PPS.  Stacking Software RAID + LVM2 + XFS is notorious for going belly
> > up on kernels configured with 4K stacks (Redhat definitely, maybe
> > others).
> 
> Really? Should I be worried? I've been running md RAID0 + LVM2 + xfs as a 
> media store partition for a mythtv box since February 2006 without issues 
> with FC4 and FC6.
> 

It depends really.  On 32bit RHEL4 (2.6.13) the thread stack size is 4K
and I have seen RHEL4 systems with 3 x (hardware) RAID5 + LVM2 + XFS +
NFS bork once a week with stack overflows.  Patching the RH kernel to
use 8K stacks eliminated this except for the occasional fubar when the
system is under very heavy load.
Much work has been done since 2.6.13 to reduce the stack usage of
stacked filesystem components (of which RAID+LVM2+XFS is just one
example) but you still see reports of "4k stack bad, 8k stack good"
especially for FC6 and FC7 which have 4k stacks by default.

For x86_64 things are different and I don't know of any confirmed stack
overflows using RAID+LVM2+XFS on such hardware.

You've been running without trouble for 18 months.  You may well run
fine for the rest of the configurations life.  Your call.
My experience of stack overflows so far has been some acceptable data
loss and a few inconvenient (Monday morning) reboots.

Have fun,
Duncan
 
> Peter
> 
> 
> This message has been checked for viruses but the contents of an attachment
> may still contain software viruses, which could damage your computer system:
> you are advised to perform your own checks. Email communications with the
> University of Nottingham may be monitored as permitted by UK legislation.
> 
> 
> _______________________________________________
> Nottingham mailing list
> Nottingham at mailman.lug.org.uk
> https://mailman.lug.org.uk/mailman/listinfo/nottingham