grp
Thu Feb 22 07:20:11 CST 2007
Thank you. For others who are looking for similar info:
Using Clustering for a Highly Available Web Site: An Example (Cluster
monitoring) for Microsoft.com:
http://technet2.microsoft.com/WindowsServer/en/library/af46359f-3712-4208-a11f-f8f1fa3e96d91033.mspx
Windows NLB does not have the capability to monitor service health.
Microsoft uses HTTPMon and ClusterSentinel from the Windows 2000
Resource Kit to monitor IIS health for the Micosoft.com website.
Probes are used every 30 seconds. HTTPMon runs through a series of
remote application level tests and collects information from each
server every 30 seconds, looking for failures in the application
layer, which are more common than system failures, but much harder to
detect. If HTTPMon determines that a server is having problems at the
application layer (for example, with IIS), HTTPMon directs Network
Load Balancing to remove the host from the cluster. One minute later,
HTTPMon checks again to see if the server has recovered. If it has,
the server is returned to service.
On Feb 22, 7:33 am, "Consultant" <consultant_mc...@yahoo.com> wrote:
> lookup httpmon
>
> <grp.a...@gmail.com> wrote in message
>
> news:1172089663.090010.203610@k78g2000cwa.googlegroups.com...
>
> > On Feb 21, 10:57 pm, grp.a...@gmail.com wrote:
> >> I'm using Windows NLB to cluster IIS servers and is working fine. NLB,
> >> however, blindly forwards traffic to port 80 even when IIS is down on
> >> one of them. The only way to avoid it handling request is by doing an
> >> "wlbs stop" to remove an affected server from the cluster to avoid
> >> traffic to an affected box.
>
> >> Is there a way NLB can be used to do health checks before forwarding
> >> traffic?
>
> > Microsoft apparently uses 8 NLB clusters of 10 servers each.(80 IIS
> > boxes) for the Microsoft.com website. How do they manage load
> > balancing w/o health checks? Is the idea to catch problems using IIS's
> > health monitors (pinging, worker process recycling etc) for self
> > correction and not depend on the load balancing solution to route
> > intelligently to responsive servers? How do you do this?