redundant OpenSwitch servers with Veritas Cluster?
Marco Stroech
Is it possible to set up two redundant OpenSwitch servers with failover via
Veritas Cluster? I mean when the OpenSwitch on the primary host fails the
Veritas cluster starts the OpenSwitch on the second host and switches all
connections over.
We would like to run ASEs with MSA replication and automatic connection
failover. Only the RepConnector seems to be able to handle with MSA, isn't
it?. Any other ideas and suggestions?
Thank you in advance.
Kind regards,
Marco Stroech
Marc Zampetti
running, and put two query lines in the interfaces file for each
OpenSwitch, one line pointing to each line. In the event of an
OpenSwitch failure, Open Client will automatically try each query line
until it gets a connection. If both OpenSwitches are configured the
same, you will get a "seamless" failover.
As for using RCM with MSA, I don't think that will work. Ideally, with
MSA, you can simply let the connections fail over. The danger here is
that some cons will be writing on one db instance, and some on the
other, and get collisions or silent overwrites. You could write a simple
CM module that directs all connections to the "master" db instance.
There is no need to do any "switch active" type work with MSA, since
bi-directional replication is setup already.
RepConnetor is a tool for using Replication Server to transform data
events into JMS messages for integration with other message-based
software, like J2EE apps, etc.
Marco Stroech
The two query lines in the interface file only works if you run applications
without jConnect. Unfortunately, nearly all of our important applications
connect to the databases via jConnect. And as I know it's not possible to
configure two servers in jConnect except you use the HA companion option
(what we don't do).
Meanwhile, we had a Sybase consultant in our house and discussed several HA
architctures with him. And it seems that there are no adequate solution when
you need automatic connection failover with jConnect and two or more ASEs
with data redundancy except we write an own RCM.
Thanks, Marco
"Marc Zampetti" <zamp...@aol.net> schrieb im Newsbeitrag
news:42adf3db@forums-1-dub...
Marc Zampetti
Yes, jConnect does complicate things as you say. You can get the
"multiple interfaces" effect if you setup jConnect to use LDAP to get
the connection information, I think. Another option would be to put the
Open Switch servers behind a VIP or DNS Rotor. That way, the VIP would
load balance across the Open Switch entries, deal with the fact that one
of the hosts is down. I'm not sure if a DNS rotor would work there.
It does sound like you'll need to write your own CM too, to get some of
the logic that you want, reguardless of how you handle the jConnect issues.
Marco Stroech
Marc,
Thanks again for your hint. LDAP is one of our future projects. I like the
idea with the VIP/DNS rotor very much and will try to setup an test
environment in a period with lower work load.
To write an own CM would be the best way probably. I doubt that we're able
to do that due to our little ressources and the forgotten knowledge to
program in C, unfortunately.
Marco