Hello,

Where can I get information and overview about webMethods EAI Clustering. I search through the “advantage.webMethods” site but it all talks about B2B Clustering.

Can someone help?

Best…

Ed,

I am not sure what version of wM EAI you are asking about. I am guessing 6.0, which I have no info about. However, up until wM ES 4.6 clustering was done using hardware. Set up two identical machines in a hot-cold configuration and use a hardware clustering solution to detect a dead instance and swap to the second instance. This only addresses failover and not load balancing.

Oh, and put your enterprise server log files for queued guaranteed documents/events on a raid drive which is used by both instances in the cluster.

Andreas

Ed, our company has two Solaris servers in Production running Enterprise server 5.0.1.
We use Sun Cluster for clustering. webMethods provides a document to setup clustering for Enterprise Server.
Both our servers are active (i.e. Active / Active setup). This allows us to make use of both Production servers. If a broker server needs to fail over it fails to the alternate machine. The shared data files (eg. Guaranteed message files) are stored on a global directory accessible by both servers. Obviously there is a lot more to this
It’s a slick solution (better than leaving one machine passive and not doing anything…)

What are you running you Enterprise server on ? ( O/S ? )
If you are using Solaris then I know webMethods provides detailed setup instructions for clustering (using Sun cluster). I would imagine that they provide similar instructions for other platforms.

Regards,

Wayne

Would you care to elaborate on this? I am interested in how you can run it hot-hot or active-active as you say. Adapters can only connect to one broker on a specific enterprise server instance on a specific machine. If you are running both active then some adapters connect to one of the brokers and some to the other. What happens in a failover situation? Are the adapters colocated on the same machine as the server and they are the ones that are started on the other machine in a failover situation? I assume both brokers are members of the same territory?

RTFM is an acceptable answer except I don’t have access to the 5.0.1 manuals, so a brief synopsis would be great.

Thanks,
Andreas

The same broker is not active on both machines, however both machines are up and running (that’s what I mean by active-active). We do balance the broker servers between the two machines (eg. Broker server A, B, C run on machine 1, and Broker servers D, E, F run on machine 2).
This means we don’t have a machine sitting there doing nothing. If a broker fails then it will fail over to the other machine. (eg. if Broker B fails on Machine 1 then it will fail over with all of it’s adapters to Machine 2…likewise if Broker F fails on Machine 2 it will fail over with all it’s adapters to Machine 1).
Get the picture ?
Each machine has a local install of the software (it is not shared). So the tradeoff is we have to apply software upgrades/service packs to both machines to keep them in sync. However this is easy to do since we just manually “fail” the broker server to the alternate machine, do our upgrade and then fail back. This way the broker server is never unavailable (unless of course both machines drop throught the floor…then we move to the disaster recovery site…knock on wood this never happens )

Here’s a blurb from the webMethods Sun Cluster manual (if you are a webMethods customer I’m sure they would send this to you):

When a client makes a request to a Broker, the Broker handles the request much the same as in an non-clustered environment. Although, in a clustered environment, the Broker
writes the client information to a shared disk instead of a private data store.
If a Broker fails after a client makes a successful request to it, the cluster software redirects
subsequent requests for the session to another Broker in the cluster.

Clear as mud ?

Wayne