MSCP_SERVE_ALL versus set device/served

[MarkOfAus]

Hi all,
A question, is there any benefit in not setting MSCP_SERVE_ALL to 1, instead leaving it as 0 & doing a set device/served for each of the devices requiring serving? This is presuming not all disks in the servers need to be shadowed.

Cheers
Mark

[Simon Fedele]

If you wanted to restrict one cluster members local disk from being potentially mounted on another cluster member, for whatever reason until set device/served is issued then I suppose that is a benefit. Our clusters have it set to 1.

[Steven Schweda]

I had mine set to 1, but I eventually decided
to change it to 4, because having the CD-ROM
drive served was causing some problems. I
normally serve all the real (hard) disks, so
automatically serving the system disk was ok
with me.

As with many things, there's more than one
way to get the desired effect.

[Con Stelios]

Simon,

The reason was primarily because I am not shadowing the system disks over the cluster (I may shadow them locally). Plus there are other disks users want to keep for isolated usage on individual nodes (application licensing etc).

I was wondering if it saves on resources and/or has "issues". Steven indicates issues with CD-ROMS.

[Dean McGorrill]

I usually serve all hard disks as well.

>to keep for isolated usage on individual nodes (application licensing etc)

a system management issue, tell the users
not to use certain area/drives.

[MarkOfAus]

.

[Robert Brooks_1]

Steven wrote . . .

I had mine set to 1, but I eventually decided
to change it to 4, because having the CD-ROM
drive served was causing some problems.

--

What problems?

[Steven Schweda]

> What problems?

It's been a long time, but as I recall, I'd
do something like mount a CD-ROM somewhere
(without saying /CLUSTER), and then I'd try
to mount a duplicate CD on a different
cluster member and get a complaint about
already having a disk with that label
somewhere. That sort of thing. I wasn't
bumping into OS defects, I just found that
not having the CD-ROM drives served made for
fewer problems in my day-to-day operations.

[Ian Miller.]

one issue I have seen is that the CDROM device has the same name on each system e.g DKA400. (Depending on how your systems are setup). This causes problems if the name becomes cluster wide.

[Steven Schweda]

> [...] the CDROM device has the same name
> on each system e.g DKA400.

Eh? Mine have names like node$DKAxxx, where
the "node" part is different on different
nodes. What does one need to do to make this
not happen? I also tend to define a logical
name or two (CD1 for the obvious drive, CDR
for the CR-writing drive (which is usually
external)), so I seldom use the real
(node-specific) device name.

I'll admit to being annoyed occasionally when
I do a "SHOW DEVICE DKsomething" and get
results from boxes all around the room, but
that happens less this time of year, and the
annoyance is more with myself than with the
OS.