My Selenium Grid setup is as follows (all VMs)
VM1 - Windows 7 x64 Windows XP x32 - Grid Hub + 2 RCs registering the default *firefox environment
VM2 - Windows XP x32 - 2 RCs registering the default *firefox environment
VM3 - Windows XP x32 - 2 RCs registering the default *firefox environment
I'm happily using Mbunit and Gallio to drive the Grid, but my problem is that sometimes the Grid hub will stop passing executions over to 1 or more of the RCs, despite their showing available on the hub console. They seem to be happily maintaining their heartbeat back to the hub, but they're never asked to do any more work. This is after they had been executing tests earlier in the test run.
Does anybody have any ideas why this should happen? In every case I've observed this behaviour, the last test an RC executed, before it then seemingly gets ignored by the hub, passed, and the session was successfully closed. Interestingly, whenever it happens to more than 1 of the RCs, its always (so far) been the pair that are running on the same VM. Yet they're managing to maintain their heartbeat, so it isn't a network connectivity problem.
Any help would be greatly appreciated!
UPDATE: After further attempts to troubleshoot, I've now discovered something else. To begin with I swapped out my Win7 VM for an XP based one, for the sake of consistency and to rule out any issues that the hub might have running on the newer OS. I've updated the above to reflect that change.
After that change, the hub is now consistently only ever using at most 5 of the 6 registered RCs. Which RC isn't used is consistently one of the two on the machine that is the last to register with the hub. So if I start the hub on VM1, start the RCs on VM2 and VM3, and then start the RCs on VM1, one of the two RCs on VM1 will never be used.