standard internal load balancer remove member from backend pool

[Harry850]

we have a couple of NVAs deployed in a backend pool of a standard internal Load Balancer.  There is a slight issue with one of the NVAs, but the health check is showing it is healthy and still serving traffic to it.  The vendor is recommending taking some logs and rebooting.  Is there any benefit in removing the NVA from the backend pool before rebooting?  Does the LB continue to send existing session traffic to the NVA until they are complete and drain away, whilst only serving new traffic to the other NVA in the pool?  If not can this be configured?  Or are all sessions terminated immediately on the NVA removed from the pool, in which case, would it be just as well to reboot the NVA without removing from the pool first?

[#2]

Hey Harry; directing traffic away from the unhealthy node during the reboot is going to depend a lot on the interval time that you've defined for the health probe.

As per the Microsoft Docs (https://docs.microsoft.com/en-us/azure/load-balancer/load-balancer-custom-probe-overview) you can specify a custom Interval Time for the health probe, which governs how frequently the load balancer will check the health status of each node in the backend pool. There is then a time-out period to consider as well - the example given is if the interval time is set to 5 seconds, you could potentially see up to 10 seconds of traffic being sent to the unhealthy node until it's marked as unhealthy in the load balancer.

To ensure there are no user sessions remaining on the troublesome NVA, the safest option would be to remove it from the backend pool and then restart.

Anthony