there weight in gold. they are using dual logging and are keeping a COB site
in parrell with their primary site. With both of these in place it would not
be too hard to recover a downed queue manager. There are obvious othere
things that can be doen like Shared Queues.


                                                          bobbee






From: Michael F Murphy/AZ/US/MQSolutions <[EMAIL PROTECTED]>
Reply-To: MQSeries List <[EMAIL PROTECTED]>
To: [EMAIL PROTECTED]
Subject: Re: MQClient Case Study
Date: Mon, 28 Oct 2002 00:02:31 -0700

What you are doing should work, however, it is not bullet proof.  Having a
client "fail over" to another server is fairly easy to do as long as your
application knows how to recognize a failed connection and to attempt to
reconnect.  You can use a client channel table to help do this
automatically, but if you use Java, you have to do it programmatically
because it does not use the table.

There is something to consider about your messages.  It sounds like you
need a guarantee that your message was processed on the zOS machine.  It
appears you have two separate MQSeries servers that will be running
concurrently and handling traffic from two different Sun machines.  If a
message is put on one, but not processed, then your client fails over,
that message on the primary server is in limbo and you don't know what
happened to it.  If it is persistent, it will get processed later, but
what do you assume in the meantime?  If you have true HA with your two
servers, you'll want to fail over the queue manager will take the
persistent messages along with it to the secondary server.  I don't know
how this is done in the mainframe world, but there are OS clustering
packages available on the others that do this.  This sort of HA works well
and should suit your solution.  With a good HA solution, your hostname or
IP will over to the secondary machine.

There are many ways to approach this so wait for more people to put in
their suggestions.

Mike Murphy
Sr. Middleware Consultant
MQ Solutions, LLC
http://www.mqsolutions.com



Idamar Ferreira <[EMAIL PROTECTED]> wrote:
Date Recieved:

10/25/2002 01:45:14 PM
To:

[EMAIL PROTECTED]
cc:


Bcc


Subject:

MQClient Case Study

Hi all,
could someone comment that topography of solution using MQClient?(See
attached file)
We'll use 2 Sun machines, running 01 IWS(Http Server) and 01
IAS(Application Server)
each one. Those Sun Domains will put messages directly in a MQServer in a
z/OS.
We plain replicate the messages in z/Os using MQ lists.
Every dotted conections will be second way to conect, not ever conected.
OUR BIG PROBLEM IS:
THE SOLUTION NEVERS CAN THINK ITS MADE A PUT AND IT IS GARANTEED, BUT
THE MESSAGE COULD BE LOST. WHEN THE APPLICATION IN IAS RETURNS TO
IWS THE CONFIRMATION, I HAVE TO GARANTEE THE MESSAGE IN Z/OS.
Our questions are:
What kind of problem We could experiment with that solution?
What about availability?
What can occurr when the MQServer(z/OS) get down? What can I do to avoid
problem and
keep the service?
If I use Name Server, Do It can solve the problem when the connection get
down, will my Client
 restart the connection in the next IP address provide by the DNS?

TIA
Idamar Ferreira
Especialista - GEROP - CAIXA
mailto:idamar.ferreira@;caixa.gov.br
((61)414-4935
"Take care of your People, They are the unique unrecoverable piece of your
business"
<<MQClient.doc>>
<< MQClient.doc >>

_________________________________________________________________
Get faster connections -- switch to MSN Internet Access!
http://resourcecenter.msn.com/access/plans/default.asp

Instructions for managing your mailing list subscription are provided in
the Listserv General Users Guide available at http://www.lsoft.com
Archive: http://vm.akh-wien.ac.at/MQSeries.archive

Reply via email to