T.R | Title | User | Personal Name | Date | Lines |
---|
2660.1 | lets work together... | OSITEL::rtont2.rto.dec.com::rtoms1::schaefere | SI-Office | Tue May 06 1997 12:09 | 12 |
| Hi Charly,
I have no answer yet but similar questions. My customer wants to operate
two MTAs (no MTA set) in a Alpha VMS Cluster (2 nodes) with transport
TCP/IP.
I know that the DSA can only be active on one node at a time but that the
MTA can run on both nodes in the cluster simultaneously.
The details, how to set up the DSA and MTA und the UA (ALL-IN-1 V 3.2)
I do not know at the moment.,
regards, Erich
|
2660.2 | No experiences with MTA sets (.0)??? | ATZIS3::EHRLICH_K | Never met a Lady like her before! | Wed May 07 1997 11:22 | 28 |
| Hi,
after looking/trying/playing around yesterday in office, I found
that
� MTA sets are only good for the MTA itselve (performance,realability)
because they can split the mail-load via 'round-robin'.
And this is a great benefit if acting as a area-'router' or a
boundary server.
� in the MAILbus 400 architecture slides: The UA has to be sure that
'he' can contact all members of the MTA sets.
-> So, from the UA-point-of-view: This is really no benefit.
YOU have to make sure, HOW to split your UAs to the members of the
MTA set. So, you can ALL-IN-1 UA connect to member1, XMR UA to
member2, maybe another ALL-IN-1 UA to member3 but, that'S all.
No way to have it automatic.
Guess, what happens if one member fails....No failover...
Or did I miss something?
Cheers
Charly
|
2660.3 | | IOSG::MARSHALL | | Mon May 12 1997 13:56 | 26 |
| A few comments:
re .1: the MTA can run on both nodes in the cluster simultaneously
This is not really a supported configuration. You are only supposed to run one
MTA on one node in a cluster. You can make it work on more than one node, but
it's not recommended.
re .2: yes, you are correct. The MTA's client interface (XAPI) available to
user agents does not provide an automatic mechanism for connecting to multiple
MTAs (in an MTA set or otherwise). In the simple case, all clients on one node
connect to one MTA (on the same or a different node) as defined by the MTA_NODE
logical. You could modify this so that different clients 'see' different values
of the logical and hence use different MTAs, but as you observe, this is a
manual operation.
You are also correct to observe there is no automatic failover between MTAs in
an MTA set. You can set-up a standby MTA so there is minimal manual
intervention (and hence time) required to bring up a standby MTA in the event
that the primary MTA, or its host node, fails, but that's about it.
If you want more automation of load sharing and failover/recovery, it's probably
best if you submit a formal product requirement to the Product Manager, Geoff
Oades.
Scott
|
2660.4 | Thank YOU, Scott! | ATZIS2::EHRLICH_K | Never met a Lady like her before! | Mon May 12 1997 15:36 | 1 |
|
|