I have two webservers in a cluster serving a web application.

Using MS Web Deploy to push a new version of the application to one server, and then again to synchronize the files to the other server in the cluster. It seems to be the most ordinary thing to do.

But wouldn't there be a problemm, when one server is deployed with the new version, and the other is not yet finished. Will it not cause troubles, when a page loaded with the new version makes a webservice request and the balancer sends the request to the server with the old version?

What's the best way to avoid this?

I thought about scripting a drainstop of the server, that we deploy to, and make sure only one server is running at a time. But I can't find anyone else, who seems to have written about such a solution. And guess that it doesn't scale very well too.

Another solution could be to shut down all servers when updating. But that doesn't seems very clever.

Any suggestions?