Has anyone experienced the following:-
GS140 ( 8400 )with 6 x 625Mhz CPU's (EV56), with 6 x BA370's attached
via 2 pairs of HSZ70's running HSOF 7.0
Digital UNIX 4.0d with patch kit 3
Users complained recently that the server was running very slowly. On
examination , I noticed that the /sbin/update
process was at 95%+ for at least 20 seconds, then faded away then
kicked in at 95%+ again when update re-runs
sync ( as set , every 30 seconds ). I killed update ( when it was
sleeping ) and ran some sync's manually.
Each sync was taking about 20 seconds to complete.
I then did some dd'ing around various filesystems, to get a feel for
I/O performance. All I can say about this was
that I was getting some very inconsistent numbers back. Some
filesystems took about 2 seconds to copy a 25MB
File , and some filesystems were taking 15seconds+.
I then ran "defragment -vn" on all of the 28 advfs file domains. Seven
of them returned with some very strange figures.
See below.
ONE OF THE GOOD ONES
defragment: Gathering data for domain 'hsz10d207'
Current domain data:
Extents: 305
Files w/extents: 29
Avg exts per file w/exts: 10.52
Aggregate I/O perf: 98%
Free space fragments: 51
<100K <1M <10M >10M
Free space: 100% 0% 0% 0%
Fragments: 36 2 7 6
ONE OF THE BAD ONES
defragment: Gathering data for domain 'hsz9d0'
Current domain data:
Extents: 4072
Files w/extents: 1457
Avg exts per file w/exts: 2.79
Aggregate I/O perf: 100%
Free space fragments: 3869
<100K <1M <10M >10M
Free space: 2146867472848995% 4% 0% 311%
Fragments: 3642 220 0 7
I then backed up one of the suspicious domains to another fresh volume
using vdump | vrestore,
and ran degragment -v on it.
After one hour it completed, and the results were still the same i.e.
...
<100K <1M <10M >10M
Free space: 2146867472848995% 4% 0% 311%
I am now getting a bit stuck on what to do next.
Bill Boddy
UNIX System Administrator
Acxiom Ltd.
0191 515 4348
bboddy_at_acxiom.co.uk
Received on Tue Mar 23 1999 - 13:36:22 NZST