Question

I’m running a command line program (happens to be Redis) inside a Windows Azure Worker Role using ProgramEntryPoint as follows

  <WorkerRole name="Worker" vmsize="Small">
    <Runtime executionContext="limited">
      <Environment>
        <Variable name="ADDRESS">
          <RoleInstanceValue xpath="/RoleEnvironment/CurrentInstance/Endpoints/Endpoint[@name='Redis']/@address" />
        </Variable>
        <Variable name="PORT">
          <RoleInstanceValue xpath="/RoleEnvironment/CurrentInstance/Endpoints/Endpoint[@name='Redis']/@port" />
        </Variable>
      </Environment>
      <EntryPoint>
        <ProgramEntryPoint commandLine="redis-server.exe" setReadyOnProcessStart="true" />
      </EntryPoint>
    </Runtime>
    <Endpoints>
      <InternalEndpoint name="Redis" protocol="tcp" port="6379" />
    </Endpoints>
  </WorkerRole>

So far, so good (it works).
I now want to run a slave instance of the server in another WorkerRole

<WorkerRole name="SlaveWorker" vmsize="Small">
    <Runtime executionContext="limited">
      <EntryPoint>
        <ProgramEntryPoint commandLine="echo slaveof %ADDRESS% %PORT% | redis-server.exe -"   setReadyOnProcessStart="true" />
      </EntryPoint>
    </Runtime>    
    <Imports>
      <Import moduleName="Diagnostics" />
      <Import moduleName="RemoteAccess" />
    </Imports>
    <Endpoints>
      <InternalEndpoint name="Redis" protocol="tcp" port="6379" />
    </Endpoints>    
  </WorkerRole>

You can see I need to tell the slave server where its master is using IP address and port; something that I don’t know until Azure has allocated the network resources for that role. I’ve seen @smarx do something along these lines.

However I think there may be a couple of things wrong with this in my case

1. I’m setting environment variables in one role and hoping to use them in another – not going to work.

2. Even if the right data was available the way I need to pass it to the redis-server.exe is not recognized as a valid entry point with the echo at the beginning

---

1. Is the only way to know the runtime IP address and Port of another
   worker role via code or is there a syntax I’m missing in the config
   file?
2. If I manage to get the IP and Port, is the only way to make my
   command line work to push it to a powershell script or batch file?

Thanks for your thoughts.

Answer 1

The only way one instance will know another’s IP address will be if a.) it programmatically grabs it, or b.) the other instance publishes it to a wellknown location (e.g. table storage). In your case, it might be easiest to just have the slave role run a startup task that accesses the RoleEnvironment (via Powershell perhaps) and sets an Environment variable with the IP Address of the master. If you do this as a ‘simple’ type, I believe it will run before your ProgramEntryPoint does (blocking) and you can just use the env var in your command line there.

Couple thoughts here however:

1. How are you handling multi-instance within a role? Are you only planning on running a single instance?
2. Do you need two different roles? Why not a single role with 2 instances that decide via election which is master?