Hi All,

We are running a load balanced architecture for redundancy with IIS7. We use OpenBSD servers running RelayD for load balancing between them. Our load balancing is not to support higher volume of traffic but to support redundancy.

In case of physical hardware failure, or a node configuration issue, we simply take that node offline at the firewall and it stops receiving public traffic.
The node can still stay up, but in some situations it may be taken entirely offline, this means the IIS process is stopped.

We are running a mix of umbraco versions but most of the sites have been upgraded to 4.2.0.1 to be put onto this new infrastructure. All of our V3 sites were upgraded, and the ones that were already V4 are on mixed versions.
ASP.Net is running 3.5.

The problem we are having is that when we take a node offline, all nodes cease to publish. We have inspected the source code and determined that this seems to be because there is a for-each loop around a webservice call, and no timeout is being set on the call to the publication webservice.
This means the publish process never completes - because one of the nodes is unreachable.
The default timeout value in the web service proxy seems to be 100 seconds - which is far too long.
It would of course eventually throw a timeout and be caught by the try/catch which wraps the for-each but then this would mean that nodes later in the for-each wouldn't be published.

We suggest that a new key be added to the web config and this key be read and set in the umbraco.presentation.cache.dispatcher class.
(Where the username and password are also set)

This is a fairly important problem, since one of our nodes is currently offline and thus noone can publish - we're not really excited about having to edit the umbracoSettings file and turn of distributed calls for all the instances.

Gareth
Sniper Systems Ltd