Article 4859 of vmsnet.internals:
Ferry Bolhar-Nordenkampf <bol@adv.magwien.gv.at> writes:

>Hi all,

Hi one...

>i'm looking for documentation about SCS (system communication services,
>the lower layer used in intercluster communication). What i'm have to do
>is to send a special kernel mode AST to a process. This process should
>do some work and then return information back to me, queuing another
>special kernel mode AST to my process. This is no problem if both
>processes are running on the same cluster member, but isn't possible
>without SCS routines if they are one different members.

...err, could be...

>I know this work may be accomplished with SCS and the CLUSTER_SERVER
>process running on every member (SYS$GETJPI uses it when obtaining
>information about processes running on different members). But i have no
>information concerning SCS. Can someone give me details on how to use
>SCS routines (docs, calling conventions, code examples)?

SCS communications, Yes!  However, the CLUSTER_SERVER has NOTHING to do 
with this!

You need to first get in touch with the basics.

This can all be accomplished quite easily if you understand the SYSAP 
communication process.  I suggest you get yourself a copy of the Roy G.
Davis book "VAXcluster Principles" and read up on this.  Particularly,
the sections on MAPping.  Check out section 2.7 "VMS Implementation of
the Block Data Transfer Service" for starters.

There are several other sections which cover the declaration of the
SYSAP, connections and the SCS directory lookup.  Then take a look at 
the LISTEN, CONNECT, ACCEPT macros for the SYSAP declaration.

Some macros which correspond to the 2.7 section are the MAP, UNMAP,
FIND_RSPID_RDTE, {SEND/REQUEST}_DATA, ALLOC_RSPID, SEND_CNT_MSG_BUF,
RECYCLx_MSG_BUF, DEALLOC_RSPID, and myriad others.

What you'll need to do is setup your ACB, SCS MAP it, use some mes-
sage protocol you devise to inform the remote node, deliver the MAP-
ped ACB, and then  respond back to the originating SYSAP, UNMAP this
memory and do general cleanup.  THIS IS NOT SIMPLE KERNEL HACKING!!!
All this happens in forks stalling and restarting within the SYSAPs
you create.  Oodles and hours of fun!

And, if you find it's difficult to understand how to create one of
these SYSAPS, it's even more fun trying to XDELTA debug one of them
when you have cluster timeout issues and such to be concerned about.

I'd suggest that if you're seriously thingking of tackling this with
a SYSAP, you'd be best advised to get a copy of the source listings 
because that's one of the only places you'll find any documentation
on the programming details of this.  

If you can, I'd suggest you look at a simpler machanism.  I devised
a SYSAP communication for my SYMBOL program but then, I dropped the
whole idea because it complicated the code and thwarted some other 
enhancements I was adding.

My original SYMBOL used (and still uses) the undocumented interface
into the SMI server.  Perhaps, if you rethink your application, you
can steal the SMI idea from my SYMBOL program as has a certain other,
much to my chagrin, without as much as changing my code or comments
and calling it theirs.  The spoils of freeware, I suppose.

VAXman-                                                 VAXman@TMESIS.COM

The memories of a man in his old age are the deeds of a man in his prime.
You shuffle in gloom in the sickroom and talk to yourself until you die.
Life is a short, warm moment and death is a long cold rest.  You get your
chance to try in the twinkling of an eye: 80 years with luck or even less!
                                             - from: Obscured By Clouds -