Actually we have a very particular problem.
All our reports are refreshed every sunday and monday morning on two servers. These two servers are not configured in a standard WebI architecture, one cluster manager and one cluster node, because they are both independant and the load balancing and fail over is managed by a hardware switch.
For reports refresh the load balancing is managed by monitoring the BCA on both servers. This works fine, but what happens is that some tasks are processed twice, once on each server.
I suppose that this is due to the “scanning repository delay” and the “update repository delay” monitoring parameters defined for our BCA on each servers. These are respectivly defined to 300 and 150 on each server.
I was wondering if someone has the same kind of architecture and the same problem.
What we do, and we’re not the only people is to avoid a hardware switch at all. Simply set up two independant Cluster Managers with BCA. Have two instances of BCA on each server and point them all to the same BCA you have defined in supervisor. Set the two key parameters to sensible values - those you have on your current set up would be okay in this situation. It’s been running for a year now I think and not one job has ever run twice.
That may not be much consolation if you need to stay with the current set up. So why not reduce your update repos parameter - just to see if you can fix the problem that way? Oh and set up bo manager logging on both servers so you can see if your theory is true.
Why would you need a load balancing switch for a batch processor?
Multiple BCA’s should simply be pointed (as suggested) to the same repository. BO keeps track whether a job has been “checked out” so that another BCA doesn’t grab it.
You don’t mention what version of the BCA you are running. There were some problems in the past, but it has been pretty solid for a while.
I have synchronized manually both servers, and for some BCA all jobs are processed twice. Is it necessary to synchronize the servers automatically ? because manually I can have a less than one second delay…
But It is working fine when I shedule one job during the day, it is processed only by one server… I really don’t understand…
My BCA version is 5.0. Could you please tell me if you have multiple BCA monitored on your servers ? And what parameters do you use for your BCA.
Thanks,
Regards,
Gautier.
nb : regarding the load balancing question, we are not using the switch load balancing for the BCA but for WebI sessions, you are right.