On Thu, Nov 09, 2006 at 02:29:04PM -0800, Anand wrote:
>fine but as soon as i run iozone with 4 threads iozone -s12G -r2048k -t4
>i see the crash.
<aol>me too</aol>
OSS,OST,MGS, etc... all on one node, and running iozone on the same node.
seems very repeatable. I've tried twice and it's happend twice - occurs
within an hour or two. apart from iozone crashing and losing files then
the node appears fine (no Oops etc.).
the only error Lustre gives is:
Nov 15 21:07:06 eh15.mckenzie kernel: LustreError:
26106:0:(client.c:579:ptlrpc_check_status()) @@@ type == PTL_RPC_MSG_ERR, err
== -28
Nov 15 21:07:06 eh15.mckenzie kernel: LustreError:
26106:0:(client.c:579:ptlrpc_check_status()) Skipped 1 previous similar message
and iozone says:
Error writing block 4003, fd= 3
Error writing block 3966, fd= 3
Error writing block 4068, fd= 3
Error writing block 4121, fd= 3
Children see throughput for 4 initial writers = 0.00 KB/sec
Parent sees throughput for 4 initial writers = 0.00 KB/sec
Min throughput per process = 0.00 KB/sec
Max throughput per process = 0.00 KB/sec
Avg throughput per process = 0.00 KB/sec
Min xfer = 0.00 KB
Child 0
iozone.DUMMY.0: No such file or directory
Child 1
iozone.DUMMY.1: No such file or directory
Child 2
iozone.DUMMY.2: No such file or directory
Child 3
iozone.DUMMY.3: No such file or directory
setup is:
RHEL AS4, dual Xeon i686
% rpm -qa | grep -i lustre
lustre-modules-1.5.95-2.6.9_42.EL_lustre.1.5.95smp
lustre-1.5.95-2.6.9_42.EL_lustre.1.5.95smp
kernel-lustre-smp-2.6.9-42.EL_lustre.1.5.95
% uname -a
Linux eh15.mckenzie 2.6.9-42.EL_lustre.1.5.95smp #1 SMP Thu Sep 28 06:36:13
MDT 2006 i686 i686 i386 GNU/Linux
# mkfs.lustre --fsname=testfs --mdt --mgs /dev/hde5
# mount -t lustre /dev/hde5 /mnt/mdt
# mkfs.lustre --fsname=testfs --ost [EMAIL PROTECTED] /dev/hde7
# mount -t lustre /dev/hde7 /mnt/ost1
# mount -t lustre [EMAIL PROTECTED]:/testfs /mnt/testfs
# fdisk -l
Disk /dev/hde: 80.0 GB, 80026361856 bytes
255 heads, 63 sectors/track, 9729 cylinders
Units = cylinders of 16065 * 512 = 8225280 bytes
Device Boot Start End Blocks Id System
...
/dev/hde5 784 906 987966 83 Linux
...
/dev/hde7 1030 5500 35913276 83 Linux
and then cd to /mnt/testfs/someDir, and run iozone with args as per Anand.
I also did this:
# ssh eh17 mount -t lustre [EMAIL PROTECTED]:/testfs /mnt/testfs
but didn't use this client mount at all
please let me know if you'd like me to try some other config or if you'd
like more debugging info.
in the mean time I'll try OSS and OST on a different node to the
MGS/MDT and see how that goes.
cheers,
--
Dr Robin Humble http://www.cita.utoronto.ca/~rjh/
and APAC National Facility
_______________________________________________
Lustre-discuss mailing list
[email protected]
https://mail.clusterfs.com/mailman/listinfo/lustre-discuss
