To all,
Part of my original posting is as follows:
--------------------------------------------------------------------------------
We have several 2100's running DU 4.0a in mixed network environments.
Currently we are running one 2100 as a small NFS server in a 3 DEC, 2 Sun
environment, with the 2100 on half-duplex 10baseT connected to a Catalyst 5000
switch. One of the Sun's is a Sparc 20 running Solaris 2.5.1 - also serving NFS,
also connected to the same switch, also half-duplex 10baseT. Another 2.5.1 Sun
is the NIS master on the network.
--------------------------------------------------------------------------------
Thanks to Knut Hellebo and Harald for their information, which helped me get a
couple of steps further.
What we have learned since:
1) nfsstat - no useful information
2) It's not just NFS. The problem also exists with ftp. NFS version doesn't
matter.
3) I added a later model Tulip interface card, and ran that full-duplex to the
ethernet switch. Transfers went from 8K/s to 15K/s. Hah, big help.
4) We ran a rolled ethernet cable directly from the 2100 to a Sparc 5. No
difference.
5) It may be a Unix LSM partition problem.
Additional info: The 2100 has 2 LSM striped partitions, /u1 is vol01, which is
striped across 2 RA29's, /netopt is vol02, which is striped across 3 RZ28's.
These 2 partitions are also NFS shared. NFS and ftp with large files works fine
with /u1. It doesn't with /netopt. The only thing that somewhat worries me is
that we have 2 RZ28's and 1 RZ28M. Does striping work across the somewhat
different types? The following info was extracted from volstat.
-----------------
# volstat -g disk10
OPERATIONS BLOCKS AVG TIME(ms)
TYP NAME READ WRITE READ WRITE READ WRITE
vol vol02 490 84 23634 4566 19.7 88.1
# volstat -g disk10 -f c
CORRECTED
TYP NAME READ WRITE
vol vol02 0 0
# volstat -g disk10 -f f
FAILED
TYP NAME READ WRITE
vol vol02 0 0
# volstat -s -g disk10
OPERATIONS BLOCKS AVG TIME(ms)
TYP NAME READ WRITE READ WRITE READ WRITE
sd disk10001-01 220 48 7982 1782 24.6 78.8
sd disk10002-01 215 18 7964 1392 20.1 92.1
sd disk10003-01 194 18 7688 1392 20.6 110.7
------------------
On this system disk10001-01 is rz3 - DEC RZ28M (DEC 0466)
disk10002-01 is rz4 - DEC RZ28 (DEC 442D)
disk10003-01 is rz5 - DEC RZ28 (DEC 442D)
I thought I had seen something at one time about LSM and different disk types.
I searched the archives with no luck. Are the different disk types causing me
problems? I'm not seeing any SCSI errors anywhere. It sure seems to affect
performance though.
Bryan
Received on Fri Oct 03 1997 - 01:03:29 NZST