we are trying to write a service for windows and need to come up with a plan for redundancy, so that if something in the application fails, it is brought back up again. i’m wondering if i can use multi-threading to accomplish this.
my idea is to create two threads which each handle separate tasks. i would like to also have each thread monitor the other thread to make sure its still running, if its not, then it should start up a new instance of that thread.
does this sound feasible? what threading techniques would i use: mutexes, shared memory, semaphores, etc…?
if this is not the correct approach, then what might be, just write two separate services and use IPC?
Well the easiest way to get your service to restart on failure is to let Windows do it. You just configure the service to auto restart, its really easy. You can also do this programatically durring the service installer. For a guide on how to do this, see this post: Building a Windows Service – Part 4: Extending the Service Installer.
As for this providing “redundancy”, this does not. The proper definition of redundancy would mean that you have more than one of them. This could be using multiple services, or using multiple services on multiple hosts more likely. Having multiple services on a single host is easier since you can use a mutex to synchronize if needed.
If your service is polling for tasks, say from a database or message queue, then the synchronization is taken care of for you. Just model your database such that multiple services can all run and process work independently without duplicating work. Now you have redundancy.