T.R | Title | User | Personal Name | Date | Lines |
---|
2687.1 | Simplistic view... | IOSG::MARSHALL | | Wed May 21 1997 12:31 | 13 |
| ALL-IN-1: the default installation is cluster-wide, nothing special to do here.
DSA and MTA: as you have a common system disk for the two nodes, you should just
be able to install everything once, in a place visible to both systems. All (?)
you then need to do is make sure that the startup-stuff is only invoked from one
machine, either by moving it to a node-specific disk or inserting checks for
nodenames, etc.
For the failover, you either just need to invoke the start-up manually from the
other machine (if the start-up is still in a common directory), or have a
separate copy of the startup stuff on the second node's node-specific disk.
Scott
|
2687.2 | I don't think it's that easy... | SNOFS1::JOYCEJENNY | Still crazy after all these years | Thu May 22 1997 01:32 | 15 |
| I don't think its that simple to fail the DSA/MTA over. At least one
of them has the node's NSAP address somewhere in the configuration....
This problem of failover has been discussed elsewhere (I can't remember
if it was this conference or the MAILbus 400 one or the X.500 one), and
I suggest you look for those discussions
Hopefully someone who knows more about this can make some more
suggestions. Personally I decided to get a separate system for the DSA
and MTA (performance considerations also entered into that decision)
and I'm currently awaiting the hardware, so I haven't actually set it
up yet.
Jenny Joyce
|
2687.3 | I hope, Internet Cluster in the solution | OSITEL::rtont2.rto.dec.com::Erich | SI-Office | Thu May 22 1997 11:06 | 9 |
| Hi all,
thanks for your inputs.
I hope I will have ONE presentation address via UCX - Internet Cluster.
Therefore agents and peer MTA and DUAs schould always "see" the same
MTA and DSA.
I also suggested to use a separate system, but the customer wants to have
all components on the cluster.
regards, Erich
|
2687.4 | | IOSG::MARSHALL | | Thu May 22 1997 12:02 | 10 |
| Using MTA sets and failing over to an MTA on a different node in the same
cluster is the most common way I know of doing this; you shouldn't need an
entirely different system.
Yes, network addresses do change between the two nodes, which is one of the
reasons I pointed out you need two different sets of startup scripts. But the
use of MTA sets should make the address change transparent to the rest of your
routing domain.
Scott
|
2687.5 | | FORTY2::PALKA | | Thu May 29 1997 11:36 | 8 |
| I believe that the UCX cluster alias will not work for the DSA, as an
inbound connection may be directed to the wrong node of the cluster.
(I would like to be shown I am wrong !).
You can define the DSA's presentation address to have multiple nsaps,
one nsap for each node of the cluster.
Andrew
|