HP3000-L Archives

January 2001, Week 2

HP3000-L@RAVEN.UTC.EDU

Options: Use Monospaced Font
Show Text Part by Default
Show All Mail Headers

Message: [<< First] [< Prev] [Next >] [Last >>]
Topic: [<< First] [< Prev] [Next >] [Last >>]
Author: [<< First] [< Prev] [Next >] [Last >>]

Print Reply
Subject:
From:
Sletten Kenneth W KPWA <[log in to unmask]>
Reply To:
Sletten Kenneth W KPWA <[log in to unmask]>
Date:
Fri, 12 Jan 2001 16:32:56 -0800
Content-Type:
text/plain
Parts/Attachments:
text/plain (65 lines)
John adds:

> I agree with Ken that Mirrored Disc/iX is a great product.
> It is extremely reliable and the performance is excellent.
> One caveat I would offer, however, is that if your system
> ever goes down with anything other than "Shut 6" (e.g., a
> system crash), it will rebuild all the mirrors upon
> restarting the system.  If you have lots of mirrored pairs
> this can take several hours, during which your data is not
> fully protected and system performance is potentially
> impacted by all the I/O going on.

Good point;  that everyone running MirrorDisc/iX needs to
be aware of:  When we first started running MirrorDisc five
years ago, I'm embarrassed to admit that quite some time
went by before I learned this:  In several cases when we
rebooted in those early days we did not get SHUT 6 for one
reason or another, and I noticed the activity lights on all the
mirror drives went nuts for a long time after we came back
up....   Although at least in our case (fairly lightly loaded;
with plenty of memory) we did not notice any serious
performance problems while the mirrors were in re-sync.


> ....  The implications of this are (1) never get impatient and
> halt your system during the seemingly infinite interval that
> it sits at Shut 5, and  (2) Mirrored Disc/iX may not be a
> good solution for systems with lots of disc. .....  when we
> had about 30 mirrored pairs.  The rebuild of the mirrors
> took about 12 hours with that configuration.

Yup....  when you are talking about hours to re-sync, then
you're starting to get some real exposure to single-spindle
failures; plus I would think pretty serious performance impact
for some sites.....   Which leads me to bring up associated
issue:


For our last couple years on MPE 5.5, failure to get SHUT 6
on gracefully shutdown was pretty rare...  actually, very rare.
And we have an APC UPS powering our 959 that will carry it
through a power outage up to 60 - 70 minutes...  But of
course if it's the weekend and none of us are around, any
extended power outage results in a crash.

Our UPS or NT servers running PowerChute software can
send an email or FTP a file to indicate when we are getting
close to running out of battery power.  Some time ago I
wrote a script that does as much of a gracefully shutdown
as it can;  that could be executed on notice of impending
power failure (turn off our background jobs;  abort users;
do CHANGELOG on our active LOGID;  LOG logid, STOP).

But without the capability to execute a programmatic
<CTRL>-A  -  SHUTDOWN, apparently will always have to
resync mirrors when come back up after unplanned power
outage.....   I've got to think (hope) that an enhancement to
allow launch of a SHUTDOWN sequence in a job would not
be a huge project....

Without a programmatic SHUTDOWN feature, I agree that
at some point the resync time becomes hard to swallow.....

Ken Sletten

ATOM RSS1 RSS2