We have 3 servers in our EpiServer environment. 1 Editorial server on the corporate network and a pair of load balanced web servers on the internet. Everything works fine with the caching events triggering content changes from the editorial server to the front end servers. We use Microsofts DFS to publish the VPP. As I say this works without issue and changes only take a few seconds to hit all servers.
The issue we see is if we reset (iisreset) web1, after about 90 seconds web2 seems to go offline or maybe re load cache? This is causing an issue since if we reset on of our front end web servers, it causes the 2nd server to timeout on page loads for up to 60 seconds and our load balancer spots this and hence redirects all traffic to a maintenance site as it thinks both front end servers are down.
I cant workout why when one of the load balanced servers restarts it causes the 2nd server to 'hang' for long enough for the load balancer to take it offline. If I just stop a server with iisreset / stop it is fine, its the starting of the site that seems to be causing the issue. Does a starting of a server cause the cache to clear or something simlar within episerver?
Strange indeed. The two front-end servers should be unaware of each other. Do you have shared session state in memory on one of the servers that causes this? What cache invalidation protocol/binding do you use on the front-end servers?
Is this also the case if you restart the editor server? Or just between the front-end servers? Does it matter which server you restart?