SUMMARY: 2 different machines on 2 SCSI bus (long)

From: Alex Harkema <HarkemaA_at_vertis.nl>
Date: Tue, 25 Jul 2000 12:18:50 +0200

Hi All,

Those are the reactions I received.

I didn't want to actually setup a cluster environment. I just want to boot
machine B if machine A is dead. So both machines would never be running at
the same time. If I want to let both machines boot from an OS disk in a
RA3000/8000 array (if possible) wich part is member specific?
I guess that the exact configuration I intended to setup is possible, but
It's a great deal of work and it will possibly (I'm quite sure)
not be supported.

----
Uwe Richter:
> I guess this creates a correct bus, right?
Yes, we've been told so too and going to order such a configuration
(3 machines + RA3000)
You have to carefully configure the system dependent settings for
swapspace, amount of memory, special devices not common to both your
DS10 and DS20 an so on
Under Tru64 UNIX V5.0 the local system disk looks like a
one-cluster-member-disk per default. Without setting up clustering
you may handcraft all references to CDSLs - but I'd guess it won't work
nor will it be supported.
You may access different SCSI-ID-identified volumes you create with
your RAID system from both machines- but you must run TruCluster if you
want to access the same e.g. AdvFS-Fileset from both machines
simultaniously. The reason is that the file system management relies
on one instance of a file set manager on one machine only (so that
locking can be save). The mass transfer will then occur simultaniously
throu the direct SCSI connection.
But for clustering unter V5.0(A) you need MemoryChannel.
(Well, GBitEthernet and/or ServerConnect-Support will be seen at the end
of the year - hopefully.)
---
alan_at_nabeth.cxo.dec.com
his is the normal configuration for an ASE environment,
	and probably supported on V5.0A as a shared storage
	interconnect for clusters.  Since an ASE environment seems
	to be what you want, you should track down the documentation
	and SPD for it and buy the supported solution.  I'm sure
	you can do it by hand, but that would probably more risky
	and complicated than the cost savings would be worth.
	Whatever you do, don't try to mount a file system from both
	systems at the same time.
---
M selcukkaraca
Let me add something here,
There is a emerging technology called SAN (storage area network).
In this new -storage- architecture.
Alpha server        Alpha Server
                  |                |
                Fibre  HUB/ Switch
                            |
            Storage Unit (e.g. RA8000 )
In this way, 2 alphas can access the same disk region. But they cannot
access the same disk -at the same time-. i.e. you must mount the disks
regions without any conflict. But if you umount from any disk from any
machine than you can mount it to other Alpha. I am sending a doc to you. I
hope it helps.
Now and another thing to here,
As far as I know, boot procedure must be done with internal disks, which are
connetced to alpha's SCSI BUS. So you should keep the alpha'a internal disks
as the same.  This is true with 4.X but with 5.X this should be not a
problem (i.e. boot from RA8000 external disk unit).
> I guess I still need some kind of member-specific disk with the specific
> configuration per machine, right?
OH YES
> Do I need clustering software in this configuration?
NO !
----
Arturo Espinosa
Alex, I think that you will  need to install 'Trucluster available
server' in both servers, in order to access the external storage unit
(this kind of configuration is called ASE Available Server Environment).
If you don't want to deal with Memory Channel, you will need version 1.x
of trucluster suite, because V5.0 needs the MC. And I think that the
fact of having two different servers is not a problem at all.
Thanks all!
regards,
Alex Harkema

Received on Tue Jul 25 2000 - 10:22:44 NZST