Drain/admin endpoint control for Load Balancer
Many on-prem systems rely on an ability to gracefully drain traffic from a node before removing it from load balancing for updates or maintenance. While there are workarounds today for the Azure Load Balancing infrastructure (http://serverfault.com/questions/686095/gracefully-take-a-server-out-of-azure-load-balancer-drain-stop) it's not as flexible as existing on-prem services. Please add this feature.
We’re working on planning this feature.
Alvin Lou commented
I add this idea on behalf of my customer, who has this user scenario as below:
Based on the current design, https://docs.microsoft.com/en-us/azure/load-balancer/load-balancer-custom-probe-overview#types, Azure Load balancer does not support UDP health probes (only HTTP, HTTPs & TCP are supported).
Hence to load balance a UDP service, a dummy application needs to be created on the Virtual machine to mark the backend service as Healthy
Please advise if it's feasible to add one of the following options:
1) ICMP health probe
2) An option to mark the service ‘always up’ (let the user decide how to check whether the service is Healthy)
Hitesh Patel commented
Feature is needed for gracefully stop the traffic specially during maintenance and code updates.
Can you guys please add this ? ...
Hi can we please have this feature now
Performing a proof of concept on Azure, day 1 hit this issue.
Back to AWS/GCP it is.
Rahul Sarkar commented
Hi Guys, This is an essential feature of any load balancer. Need this asap to implement seamless rolling updates to VMs in a scaleset...
Jason van der Paal commented
We have waited long enough,
We're now looking at moving all workloads requiring load balancing to another cloud provider.
Azure has half a load balancer.
Alex Bieber commented
Dear Azure Networking Team. Please comment here and tell us if there is a rough schedule for providing this as it is in the feature backlog for 2 years now.
I'd like to scale down the web server without dropping the connection, but there is no function to implement DrainStop for web servers in LB's scaleset. please consider adding drain connections features such as Windows NLB manager.
Please implement this feature
Is there an update on when this might be fixed? Without this “feature” load balancers in Azure are a bit rubbish! Even the inbuilt windows server network load balancer which is ancient has this functionality.
Chris Bennett commented
Second that comment from Jeff Miles. Surely this is a must have feature for any load balancer? If it hasn't climbed to the top of your backlog by now I'd hazard a guess that it never will.
Jeff Miles commented
This feature is a must-have. Any update about it's status in the backlog would be appreciated.
Charlie Scherker commented
I have to add a "me too" to this article. Weare currently migrating our entire infrastructure to Azure and find this situation to be very frustrating. Is this still under consideration?
Julio Catzalco commented
This shouldn't even be a suggestion, but a built in feature. It's like designing a new car that allows you to get in, but *oops* we forgot to include a way to get out. 1) Maybe you can roll down the window, or 2) maybe bust it out, no wait, 3) have someone from the outside open the door for you. Yeah, it's like that.
Matthew Miller commented
Any schedule for this feature yet?
Sam Hodgkinson commented
Has there been any update on this feature getting added, into the azure load balancer.
Francis Sirizzotti commented
Currently, Azure Load Balancer does not support any way to programatically mark a node unhealthy or otherwise remove it from the pool temporarily during maintenance. Meaning you have to accept errors to your end users while deploying. It'd be great to either allow a request from the node, or a secondary health check to mark a node as unhealthy without it actually sending errors back to the user.
Mark Nash commented
Is there any Update to this as it has been almost half a year since It became under review?
We are looking to implement a way to easily update load balenced servers without taking sites down and this may be the update to Azure that we require.
This is very important we have Azure VM's serving web pages using IIS in shared config mode and using a file share as the source directory. If I use the recommended solution then both VM's would report as down. There is no way to take a VM out of rotation to perform updates.