backup and analyze/disk on volume set gets SYSTEM-F-DEVNOTMOUNT error

[Dan Mellem]

Hi,

We have a VAX cluster of three 4000/600s. There's a volume set of four RF35 disks that recently had two disks fail (trouble spinning up) and were replaced. We initialized the volumes:

init /owner=system/extens=100/nohighwater -
/headers=2500

and bound them:

mount/bind DATA_DISK $1$dia2:,$1$dia3:,$1$dia4:,$1$dia5 DATA1,DATA2,DATA3,DATA4

and then restored the backup:

backup/ignore=label mka500:.bck $1$dia2:[*...] -
/by_owner=original

Dismounted, deallocated, and mounted system. Everything seems to work fine. However, when we try to do a backup or anal/disk, we get the error:

VAX1> anal/disk $1$dia2:
%ANALDISK-F-OPENINDEX, error opening INDEXF.SYS, RVN 2
-SYSTEM-F-DEVNOTMOUNT, device is not mounted


When we did a show dev $1$dia2:/full, we discovered that RVN 2 was skipped somehow.

$1$DIA2 DISK2 DATA1 rvn1
$1$DIA3 DISK3 DATA2 rvn4
$1$DIA4 DISK4 DATA3 rvn3
$1$DIA5 DISK5 DATA4 rvn5


Is there any way to fix this problem besides a full backup (via copy?)/initialize/restore?

Thanks,
-Dan

[John Gillings]

Dan,

And people wonder why I don't like volume sets! (enter Dan's post into evidence, prosecution rests m'lud ;-)

Did you reinit ALL the volumes, or just the new ones? Maybe the surviving volumes kept their original RVN? If they were 3 and 4, then maybe the (apprently) new volume was tacked on the end? If that's the cause, you probably don't have any choice but to wipe it all and start again. This time check the RVNs immediately after the bind.

It always bugs me to use the physical name of one of the volumes to refer to the whole set. Perhaps if you use a logical name it will help? Try

MOUNT/SYSTEM/BIND=name -
disk1,disk2,disk3,disk4 -
label1,label2,label3,label4 MYVOLUME

Now try BACKUP MYVOLUME and ANALYZE/DISK MYVOLUME?

Better still lose the DSSI technology. For the price of the POWER on those large physical size, power hungry, but tiny capacity drives you could replace the lot with a single spindle with maybe 10 times the capacity! My iPaQ has more than SIX TIMES the capacity of your entire volume set on a mini SD card the size of my thumbnail!

[Ian Miller.]

"Is there any way to fix this problem besides a full backup (via copy?)/initialize/restore?"

Unfourtunally not. Seriously consider getting rid of those drives. They have not made RF series drives for 20 years so any replacements you get are old too. Even a HSD10 in a BA350 shelf with five RZ29-VA would be an improvement.

[Dan Mellem]

Hi, John,

Thanks for your answer. Yes, we initialized all four volumes and restored the backup. We also generally use the logical name DATA_DISK but let me do some more testing with it as you've suggested.


Hi, Ian,

Thank you, too. That's what I feared. We couldn't find any documented way of doing it but thought there may be some trick.

I'd love to get rid of the RF35s and the volume sets but many of the programs are expecting all of our existing volume sets (four of them) and I doubt budget would be approved to improve a legacy system. We do have an HSD10 for one set, though (4x 2GB). You're right, though, 3.2GB on four ancient disks is a bit primitive! Also, decreased reliability since there's no redundancy.

-Dan

[comarow]

You folks are missing the joys of volume sets. Where else could you combine an
RL02, a TK50, an RZ28 and just about anything in one logical volume set. That's an essential feature.

It is a cheap way to spread I/Os with no change to applications

Seriously, I wonder what is often the cost of saving money and endangering systems and applications, where new peripherals are so
inexpensive.

[Ian Miller.]

Dan,
you can work around device name depenencies with logical names. The small amount of money spend on another HSD10 and shelf can be justified in less downtime, saving on maintaince contract etc