IIS Zero Downtime Update ARR / Reverse Proxy

Question

I have a C# console application / Windows sevice that uses the HttpListener stuff to handle requests, IIS is setup to reverse proxy to this via ARR.

My problem is that when I update this application there is a short downtime between the old instance being shut down and the new one being ready.

The approach I'm thinking about would be to add 2 servers to the server farm via local hostnames with 2 ports and on update I'd start the new instance which would listen on the unused port, stop listening for new requests on the old instance and gracefully shut it down (ie process the current requests). Those last 2 steps would be started by the new instance to ensure that it is ready to handle the requests.

Is IIS ARR load balancing smart enough to try the other instance and mark the now shut down one as unavailable without losing any requests until the new one is updated or do I have to add health checks etc (would that again lead to a short downtime period?)

Answer 1

One idea that I believe could work (especially if your IIS is only being used for this purpose) is to leverage the IIS overlapped recycling capabilities that are built-in when you make a configuration change. In this case what you could do is:

1. start a new instance of your app running listening in a different port,
2. edit the configuration in ARR to point to the new port.

IIS should allow any existing requests running in the application pool within the recycling timeout to drain successfully while new requests will be sent to the new application pool.

Maybe if you share a bit more on the configuration you are using in ARR (like a snippet of %windir%\system32\inetsrv\config\applicationHost.config and the webFarms section)