[OpenAFS-devel] 1.2.10 linux kernel hang during AFS backups

[Harald Barth]

> I am having a serious problem with 3 different AFS fileservers
> hanging during AFS backups and would appreciate some help from
> someone who knows the linux/AFS internals.  This happens every
> night basically during AFS backups on one or more of the 3
> machines.

You did not provide a lot of info about your server/OS setup. But
my guess is Linux 2.4.20 on x86 ;-)

> <5> [<e0f009b0>] afs_xdcache [libafs-2.4.20-19.9-i686.

Upgrade to 2.6 kernel or a 2.4 kernel with patched memory management.
We (PDC) have the patched 2.4 running on some servers but today I
would recommend to go for a 2.6 kernel of your choice directly. You
might not be able to run the client you want on 2.6 but at least the
kernels memory management is stable enough to stand som serious AFS
server load. Or run Solaris x86 if your HW does permit that. Or
FreeBSD.

It is a pity that Solaris x86 lacks driver support for a lot of useful
things (like the onboard RAID shipped in Suns Intel boxes, Dell and
HPaq), otherwise it would have replaced Linux as the stable OS of
choice to run my AFS servers on a long time ago.

Harald.

My x86 matrix of things that should work (this week):

OS		Server		Client

Linux 2.6	OpenAFS 1.3	OpenAFS 1.3 or Arla 0.36/0.37 or none
Solaris x86	OpenAFS 1.3	OpenAFS 1.3
FreeBSD		OpenAFS 1.3	Arla 0.36/0.37 or none