AppFabric FAQ: Availability

Availability

Q: What do you offer to customers that need high availability and fault tolerance?
A: High availability of the service is provided through the instance restart service (IRS). It monitors the workflow status and resumes the suspended instances after the system or application failures. High availability of the processes (i.e. IIS worker processes) is provided by IIS. High availability of the server nodes need to be managed by enterprise tools or using custom solutions outside of AppFabric.

Q: How does this work with Disaster Recovery (DR)? Typically, the DR site services are down but enabled in the configuration, so that when a failover occurs the correct service instances are started. Replication between SQL server stores, 4 machines in each primary and DR site.
A: All services are durable in persistence store. Replicating them into another store is how they will be maintained. IRS will bring up all those services on the other machines. Then route the messages to the other machine.

Q: Are you going to manage persistence/monitoring databases in a distributed environment, and what is the story around clustering/partitioning this data?
A: In “AppFabric”, we support configuration of persistence/monitoring databases using “AppFabric” management tools. Also, we support farm scenarios where identical nodes share the same persistence and tracking DBs. We rely on SQL for clustering and partitioning.

Q: What will happen in a scenario where the workflow has started but has not yet completed. It has been persisted, and the server it was executing on becomes unavailable for a period of time. If the execution is limited to a single server does this mean the instance cannot continue until the server becomes available again?
Execution is not limited to a single server in the typical scenario - usually workflows are not affinitized to any particular server or host. If the WF has been persisted and another message arrives for it, then the WF will be re-loaded to the server that the message arrived. If the WF was running and the server crashes for some reason, then the instance restart service will be able to bring the instance back up on another machine in the farm.