Hi John,
We see this same failure from time to time - and since this is a site
with full HP Maintenance, the support engineers have been analyzing the
problem quite deeply.
They found the reason to be SCSI-bottlenecks and really didn't come up
with other solutions than adding another SCSI-card. Since there is no
more space available in the server for this, we live with the mirror
fallouts, and simply rebuild the mirrors, when this happens. (The disk
system is monitored by a batch job and an e-mail alert is sent, as soon
as one of the disks quits the mirror).
We use Backup+/IX and this is indeed a great piece of software but it
really presses the disks very hardly - and the explanation from the
engineers was this:
Mirror/iX piles up mirroring requests, but when the I/O-system is very
busy there is not enough time left to complete the mirroring. When the
stack is full, the system has to drop the mirroring and continues with a
single disk in the pair. The alternative to this would be the system
aborting.
Another issue: We had a few groups once, containing an extreme amount of
small files - and doing backup of all these files quite often led to
loss of one of the mirrored partners. Spreading these files out on more
groups solved this part of the problem.
We also do some batch reorganization and capacity changes of IMAGE
databases during off hours in the weekends, utilizing DBGeneral for
this, and of course these are very I/O-bound tasks, and gave us the same
problem. To solve this problem - or rather to minimize the number of
fallouts - we stopped all activity except one single 'serialized'
reorg-job.
We also discussed pros & cons re. having a few large disks or a bunch of
smaller disks in the mirror. No evident conclusions were made on this
issue then.
HTH.
Mange hilsner / With kind regards
Hans-Ole
===========================================================
E-mail: [log in to unmask] Cell Phone: +45 4042 5507
===========================================================
ScanConsult IT-Partners ApS Voice : +45 7734 7474
Jegstrupvej 96A Fax : +45 8738 1416
DK-8361 Hasselager Support : [log in to unmask]
Denmark Sales : [log in to unmask]
http://www.scanconsult.dk
= ParaSuite : A Suite of Air Cargo Handling solutions =
= Partners : Hewlett-Packard, Cognos, Microsoft and more =
-----Original Message-----
From: HP-3000 Systems Discussion [mailto:[log in to unmask]] On
Behalf Of John Lee
Sent: Tuesday, November 14, 2006 10:38 PM
To: [log in to unmask]
Subject: [HP3000-L] MirrorDisk question
We just added 4 disk drives to 2 Jamaica arrays and mirrored the drives
with MirroriX. All was well and we started (or I should say MPE
started)
moving data to the new drives. 24 hours later, the mirroring was lost
to
both drives we mirrored to. Anybody had a similar experience? I'm not
a
MirrorDisk expert. What's it telling us?
TIA,
John Lee
Vaske Computer Solutions
* To join/leave the list, search archives, change list settings, *
* etc., please visit http://raven.utc.edu/archives/hp3000-l.html *
* To join/leave the list, search archives, change list settings, *
* etc., please visit http://raven.utc.edu/archives/hp3000-l.html *
|