Load balancer weight setting for backend nodes in a pool
Add the option to set “weight” to load balancer backend nodes in a backendpool. This enables the usage of active/passive virtual network appliances without the usage of route table adjustment scripts etc....
Hi Kevin – thanks for the feedback, we are always looking to improve Azure LB.
— Anavi N [MSFT]
Andy Ziegelbein commented
I also voted for this feature but wanted to provide some additional feedback. Active/passive NVA deployments are needed to meet the requirements of some customers, but the use of scripts and API calls is both complex and error prone. It's not uncommon to have different teams managing the cloud-native networking environment (e.g. route tables) and the NVAs. These two teams must stay in sync any time changes are made (e.g. new routes, new address space) and this can be an issue. If the CloudOps team adds a new route and the team managing the NVA script doesn't add this route to their script, only a subset of the routes will be updated to point to the new active firewall resulting in a partial outage. This of course defeats the purpose of having an HA NVA solution.
Some NVAs are capable of establishing a heartbeat between themselves and can be programmed such that only one of the two NVAs will respond to a probe at any one time. We've done this with the Cisco CSR 1000v using EEM, but this is more of a hack. Using a weight like Kevin suggested would provide a much cleaner, more elegant, and less error prone method for setting up active/standby NVAs.