We have a setup with 2 VMs in azure, each running IIS with the same web site. The VMs are in a backend pool of an azure internal load balancer with session persistence set to 'none'.
When we do load testing and send traffic to the LB, we expected it to distribute the traffic evenly. But instead we see a strange behavior where 1 VM gets all the traffic for 20 seconds, then it switches to the other VM for about 20 sec and then back again.
See screenshot over cpu usage (looking at perf counter for incoming network traffic on the nic shows the exact same behavior).
Are there any settings on the LB that can cause this behavior?