Orleans rolling release and strongly typed grain state

[jbockle]

I'm trying to find a viable solution to rolling releases to prevent application downtime. I have a aspnetcore 3.1 co-hosted silo, it doesn't have any external cluster clients. When I do a rolling deployment now, if one instance is newer, an older instance that activates a grain on the newer silo instance encounters an exception if the grain state class has changed in some way. Once the rolling deployment has completed, this issue goes away.

Obviously, I want to prevent that from happening. I had a thought that maybe I could use cluster id or service id as a cluster version.

e.g.

builder
    .Configure<ClusterOptions>(options =>
    {
        // could I change either between builds to append a version to prevent older 
        // instances from activating grains in a newer silo instance?
        options.ClusterId = "FOOAPPCLUSTER";  // change to "FOOAPPCLUSTER_V1"?
        options.ServiceId = "FOOAPPSERVICE"; // change to "FOOAPPSERVICE_V1"?
    })

Would this be a viable solution? Are there any concerns with this approach?

[jbockle]

Nevermind, I see the documentation explicitly states you can change the cluster id between deployments so only silos with the same cluster id will communicate with eachother, https://dotnet.github.io/orleans/docs/host/configuration_guide/server_configuration.html#orleans-clustering-information