Walt replied to the list, RE my:

> Mirror Disc factoids; and stuck in "BUSY"...
>
> 1. Pulling out or pushing in a drive into the  HASS will
> trigger a bus reset which will take 30-45 seconds.

I was pretty much on track that far...

> ....  Doing that back to back will cause two bus resets
> that will exceed the 80-105 second timeout value for
> detecting a disabled mirror.

AH HA !!...  now it becomes clear....  thank you for
explanation....

> 2.  Pulling out a drive when no I/Os are pending will not
> trigger a disabled mirror.

That's reasonable;  if there isn't anything that needs to use
it at the moment, unless you did regular polling to generate
pretty constant low-level disc activity (doesn't sound like a
good idea), how would MirrorDisc know it was gone....


> 3. Without letting MPE settle down before each of the events
> that you did might (will) expose problems ouside of my
> direct influence, And I hate that.
>
> (Happy Mirroring)

Been mirroring for just over five years now:  Have never lost
a mirrored spindle (started with 10;  up to 20 now);  had only
three or four mirror DISABLED events I did not deliberately
cause during that time, IIRC.....    We be happy mirroring
campers...   :-)


hmmm.....   while have it in mind:  Walter or anyone care
to comment on latest revisit of poor-man's mirroring the
MPEXL_SYSTEM_VOLUME_SET (since IIRC one the key
sticking points if HP was ever going to try and mirror the
<long_name> was deciding which spindle to go to during
BOOT-up)...  I know HP sez no plans to implement mirroring
of SYS_VOL;  but humor me for a bit:


At the ISL prompt you have a chance to change PRIMARY
boot path.  If mirroring could handle keeping SYS_VOL
mirrored while up and running, how about:

(1)   One or more spindles in SYS_VOL blows up:  CRASH.

(2)   You attempt to BOOT in SINGLEDISC mode.  IF that
works, LDEV 1 is (probably) O.K.;  it's one of the other
spindles in the SYS_VOL.  While booted in SINGLEDISC
mode, you are (by some STM magic or etc. that we won't
worry about right now) able to determine which other LDEV
in SYS_VOL has gone south.  Even if I had to switch the
bad "non-LDEV 1" member of the SYS_VOL to a mirrored
sibling LDEV manually while booted in SINGLEDISC mode,
I would be happy if I could do that.... or at least:  Happy
enough.

(3)  If attempt to boot in SINGLEDISC mode does NOT work,
then (assume) LDEV 1 is bad.  Change PRIMARY boot path
at ISL prompt;  reboot;  pull out the failed LDEV 1 hot-swap
module;  stick in a new spare;  let mirror resync while up
and running;  buy another hot spare;  all is well....


Could something like the above make is sufficiently "easier
enough" that HP *might* re-consider that level of mirroring
the system volume ??...   Right now to get back up if any
spindle in the SYS_VOL fails I've still got to figure out which
drive failed;  *AND* do an INSTALL of entire SYS_VOL.
Something like the above would allow skipping the INSTALL.

BTW:  Did I say MirrorDisc is a great product;  some kind of
SYS_VOL assist is the only key missing feature ??...     :-)
I know a few sites have posted to the list about problems with
frequent DISABLED mirror spindles or etc....  all I can say is
sure has been running well for us for 5+ years....

Ken Sletten