This content has been marked as final. Show 1 reply
> What exactly does "stops running" mean here from the connector perspective ?
> Is it;
> * HTTP requests fail (and if so what status codes are regarded as permanent
> * TCP connectivity is lost
> * Something else ?
In my experience with this, if you're talking about the CF level of
clustering specifically, the failover from one CF instance to another
occurs when a CF request fails. This doesn't have anything to do with
HTTP requests so if your web server isn't clustered the CF clustering
isn't going to help you there.
We did a lot of analysis on this recently as we built a new
high-availability environment. The basic question you have to answer is
when a physical server is considered "down" and should be taken out of
service. At a high level you have three situations to consider:
1. Box is up and running and responding to TCP/IP traffic, but IIS is down
2. Box is up and running, but CF is down
3. Box is dead
CF clustering doesn't help you with IIS being down, and of course it
doesn't help you if the box is down. The absolute best solution overall
is a hardware load balancer because that can (depending on your load
balancer) help you at both the box level and the IIS level, and with a
bit of extra work even with the CF level.
In our situation we decided that since we have three servers, if any one
of the three situations above occurs, the box is considered out of
service and the load balancer fails over to another box. We could
cluster CF but decided it wasn't worth the effort and additional
overhead in our situation. Your needs may dictate otherwise of course.
Hope that helps,
Adobe Community Expert - ColdFusion