[Info-vax] Quorum disk and page/swap/dump files

Stephen Hoffman seaohveh at hoffmanlabs.invalid
Fri Dec 6 09:16:55 EST 2013


On 2013-12-06 13:49:29 +0000, RGB said:

> We have a 2-node HP rx2800 cluster running VMS 8.4.  As such I have set 
> up a quorum disk for the 2-node cluster with the disk having a vote of 
> 1.  (Expected votes are set to 3.  Thus quorum is 2.)

Given subsequent references to EMC, I'm guessing this is probably on 
Fibre Channel, and a Fibre Channel SAN disk would be accessible to both 
hosts.  If not and if you're using some sort of local EMC storage, then 
the quorum disk won't be useful.

> However, that is neither here nor there.  I was only able to obtain a 
> 32GB LUN from our EMC folks for use as the quorum disk.  Which seems an 
> awful waste of space.

Ayup.  Wasting an entire half-of-a-cellphone's worth storage.

> Does anyone see any problem with putting the cluster's page, swap and 
> dump files on the quorum disk?  I've never done this before and I'm 
> hesitant in doing so due to the critical nature, obviously, of the qdsk.

Will this work?  Sure.  More than a few folks use a shared boot disk as 
the quorum disk.  Some will put a bootable backup root on that disk.

This all assumes you're still paging, too.  With the increase in 
physical memory, paging is becoming less common.

The biggest general concern with the quorum disk is the responsiveness 
of the I/O aimed at the disk.  If the quorum disk I/Os get delayed due 
to saturation, then the cluster will "burble."  Excessive paging I/O 
could cause this, and disk backups have been known to cause 
quorum-related cluster connectivity glitches, too.

Would I bother with this savings?  I'd ask the EMC folks how tight they 
are on storage, and — guessing — they're probably not, and if so then I 
probably wouldn't worry about this.  Might stick some licenses or other 
common data there for general use, maybe even the cluster common files, 
but...

(FWIW, there could well be as much interstitial loss on an OpenVMS disk 
with large numbers of files and large cluster factors, if not more.  
Average loss within an OpenVMS file system is half the number of files 
on a disk times the disk cluster factor, and loss can be higher with 
some file organizations.  While I'm doing math, 32 GB is ~0.032% of a 
terabyte disk, and a one-terabyte disk is a small disk by current 
standards.)


-- 
Pure Personal Opinion | HoffmanLabs LLC




More information about the Info-vax mailing list