Re: ZFS kmem_map too small.

From: Craig Boston <cb_at_severious.net>
Date: Fri, 5 Oct 2007 08:40:17 -0500
On Fri, Oct 05, 2007 at 02:00:46AM +0200, Pawel Jakub Dawidek wrote:
> If you can still see those panic, please let me know as soon as possible
> and try to describe what your workload looks like, how to reproduce it,
> etc. I'd really like ZFS to be rock-stable for 7.0 even on i386.

Have two of them for you.  My amd64 system is offline for an
indeterminate amount of time, so I don't know what its status is.

#1)

The one I can reproduce it on the most reliably is my old fileserver.
It only has 512MB RAM, which I know is below the officially supported
limit, but it _used_ to work perfectly (even if a bit slow) with the
settings tweaked.

hw.machine_arch: i386
hw.realmem: 536309760
vm.kmem_size: 306184192
vfs.zfs.arc_max: 167772160
kern.maxvnodes: 25000

About 1.5-2 months ago it was working fine with the autotuned arc_max
value -- I don't remember exactly what it was.  Once the panics started
I've been slowly bumping it down, but no matter how low I set it, it
can't seem to survive a buildworld & buildkernel.

I don't have the ZIL disabled on this one, as I'm running PostgreSQL on
it and prefer fsync() to work.

#2)

This one panics far less frequently and is more difficult to trigger.
My latest ARC size reduction seems to have it almost stable, but with
the autotuned value would panic within the day.  Besides the kmem_map
too small panic, I sometimes get a "could not allocate kthread stack"
panic, which MAY be related.  Or it may not.

hw.machine_arch: i386
hw.realmem: 2145959936 (2GB)
vm.kmem_size: 671088640
vfs.zfs.arc_max: 300000000
kern.maxvnodes: 75000

The other odd thing about this one:

vm.kvm_size: 1069543424
vm.kvm_free: 0

I don't know if I should be concerned that kvm_free is consistently zero
on this system, but it's a bit strange as all the others I work on have >0.

Please let me know if there's any information I can provide that would
be useful.

Craig
Received on Fri Oct 05 2007 - 11:40:19 UTC

This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:39:18 UTC