Clustered Workflow Servers

[Michael_Butt1]

I'm not sure if this is the right spot to ask, but I have a customer that runs a large number of business critical workflow timers. The timers perform various tasks including running scripts, exporting data, calling web services and interfacing with external database tables. So keeping these timers up and running is important to the customer.The timers are mostly core based using the Workflow Timer Service, but here are some thick client based timers as well.

We would like to use Windows Clustering in an active/passive configuration for the servers running these timers. I believe that means that the same timers and any local resources they interface with will have to be setup on both nodes.  Is this possible?

I thought I remembered that Workflow Timers could only be applied to one timer service or client instance, but my co-worker believes it is possible to cluster them.

[Seth_Yantiss]

[quote user="Jose Pineda"]MS Clustering wouldn't be more expensive than buying the extra licenses

So yeah... For most probably...  I'm not too sure what MSFC costs, in all honesty.   Since I work for a Fortune 500 company, we have pretty much everything MS offers in some capacity or another as an Enterprise license.

[quote user="Jose Pineda"] I would be very interested in knowing if this works when theOnBase service just 'hangs' but the server is still up. 

If the Service just "Hangs" but doesn't die, Failover Clustering doesn't see it as a problem and won't fail the service over to the secondary node.  And yeah, that is what happens a good deal of the time... Just hangs on some document...  

[quote user="Jose Pineda"]Would it have been better with clustering... maybe...???

Well, there's really two problems to overcome with systems like this.

1. High Availability - Partly solved through the use of Failover Clustering and some Load Balancers
2. Dependable Performance - Partly solved through system diversity and most Load Balancers.  

You're distribution of Timer Services is a fair solution for the 2nd problem, and does a bit to address the 1st in that the impact of a single server going down doesn't take out every process at once.  So in and of itself the question "Is Failover Clustering a better option?", is best answered by weighing the cost of the product and the load of all timers on one machine against the cost of multiple servers, the impact of losing some timers for an unknown amount of time, power, network drops, etc.

For us, this was easy to answer.  Fewer servers is something of a mandate for us.

Cheers,
Seth 

[Michael_Butt1]

This is some fantastic information. It was exactly what I was looking for, thanks for the discussion and instruction.

Seth, I have a quick clarification about your WF Timer Service implementation if you wouldn't mind since this discussion branched out in several different directions. You have two Timer services, one on each cluster node. These services are configured the same, with the same timers running on each instance of the service. Then they are both set to MANUAL and the cluster software decides which one runs. Is that an accurate description?  I had assumed that you could only assign the timers to a single server and they wouldn't be available for assignment to a second server after that, but I had never tried it.

I would like to address the questions about what is more cost effective, clustering or buying additional licenses.  Based on my research for this project it appears that Microsoft Clustering is not a license in an of itself. It is a feature that is or was included on certain versions of Windows Server (Enterprise, Database and Hyper-V).  New in Server 2012, however, Windows Clustering is included for up to 2 Nodes in Server 2012 Standard. So there is no additional cost for clustering functionality.

SOURCE

[Seth_Yantiss]

Michael, Your write up is exactly correct with the caveat' that your servers would need to be set up for MS Failover Clustering and correctly configured for service failover.  Additionally, the Hyland Workflow Timer service would be configured inside the Failover Load Balancing Administrator Console.

[quote user="Michael Butt"]I had assumed that you could only assign the timers to a single server and they wouldn't be available for assignment to a second server after that, but I had never tried it.

Yup, you can actually run the Workflow Timer Service on as many machines as you like and add 100% of your timers to each service instance.  REALLY REALLY not recommended, though for a host of reasons.

Thank you for the information regarding the cost of the service!!!  I hope that's useful for anyone who's looking to answer similar questions.

Cheers,
Seth