On 09 Mar 2001 09:18:17 -0600, Steve Lord wrote:
> Can you supply some more details on your setup (mount options, roughly
> what you are doing in terms of how much data you are pushing when this
> happens etc). We can try and replicate this here.
The server is a Dual-300 MHz Pentium II on an intel 440BX-based
motherboard. There is an Adaptec 2940UW SCSI card, on which / is
(Seagate ST34501W). Root is currently ext2.
The onboard PIIX4 IDE controller has:
hda: IBM-DTLA-307075, ATA DISK drive
hdc: ATAPI CD ROM DRIVE 50X MAX, ATAPI CD/DVD-ROM drive
ide0 at 0x1f0-0x1f7,0x3f6 on irq 14
ide1 at 0x170-0x177,0x376 on irq 15
hda: 150136560 sectors (76870 MB) w/1916KiB Cache, CHS=9345/255/63,
hdc: ATAPI 50X CD-ROM drive, 128kB Cache, UDMA(33)
/dev/hda1 is formatted to be one big XFS partition and is exported via
NFS, v1 or v2 only. (We had previous problems with v3 on the clients, so
decided to just stick to v2.) NFS is from Red Hat 7.0
nfs-utils-0.1.9.1-7 -- the system is basically stock RH7 with all of the
errata applied (via Red Carpet, of course ;-), and then a kernel built
from linux-2.4-xfs CVS as of this morning.
The directory is exported "*(rw,no_root_squash)" (it's on an internal
subnet), rpc.mountd is running with --no-nfs-version 3, and we have the
kernel nfsd running. Hmm. Kernel nfsd is indeed running. Maybe I'll
try this with userspace nfsd, yes? Dunno if that would have an effect
The client is running roughly the same version of the XFS kernel and is
just doing a 'cp -r' of a directory with roughly 12 or so files, total
70MB. The last file is often the one that gets an IO error. This cp is
happening while a few other clients are reading/writing uncompressed
.wav's to/from the nfs drive. Also, if the files are copied over via
tar, it can apparently succeed, but tar has to recover from some errors.
Ian might be able to provide more information when he sees this.
Hope this helps; I'll keep poking at it on our end, and I will give
userspace nfsd a try as well (and turn on debugging perhaps).