Problem with Azure Internal Load Balancer

Jonathan Stockley

We have deployed BOSH 261.4 with Azure-CPI 260.5 with stemcell bosh-azure-hyperv-ubuntu-trusty-go_agent version 3363.27.
I have deployed a BOSH release of our own application that uses two VMs and an Azure Internal Load Balancer.

Everything works great for about an hour and then the ILB stops responding. The two VMs that are in the backend pool are still working fine and the health check probe URL is still responsive in both VMs.
However after the ILB stops responding, there is no sign that it is sending the health check probe.

I have been working with Microsoft Azure Support but so far we have not discovered any cause for this. It doesn’t help that there are no logs or diagnostics in Azure for an Internal LB, only for External LB.

As a comparison, I manually deployed two Ubuntu Trusty (14.04) VMs using the Azure marketplace and setup an ILB in front of them and it has been working for the past 3 days without a problem.

Has anyone else tried to use Azure Internal LB with BOSH?

Could it be that the Azure stemcell we are using has a problem?

Any help or suggestions would be most welcome.