IIS 7 and Above
Application Request Routing (ARR)
Managing ARR Server Farm through Monitoring and Management menu
Last post Aug 14, 2019 07:00 AM by brutlex
Aug 06, 2019 07:37 AM|brutlex|LINK
I have a working ARR + Windows NLB setup where I run multiple server farms on an active/active ARR+NLB for which I use shared configuration.
Now I have an issue and an additional minor question:
I have attempted to do a graceful drain of one server with 500+ connections in order to perform maintenance with no impact for the users. Whenever I mark a server as unavailable "gracefully" it works perfectly, but then it switches seemingly at random back
to Available after a few hours. This does not necessarily happen on all ARR servers in my group and i have not found an apparent reason for it.
The additional question which is bothering me is that, unlike most settings, the ones found inside "Monitoring and Management" menu for a Server Farm are not written to the shared config file. I am wondering where settings like server availability is stored?
Any help or info would be greatly appreciated!
Aug 06, 2019 03:36 PM|Rovastar|LINK
The reason for servers suddenly reappears is because the application pool is recycled and it defaults to how you setup the farm (where the servers online or offline when you set it up). It is safest to take these servers offline in the server section (as
well?) then it is written to the shared config. You should really use the monitoring and management for that. I know it implies you can do this and looks logical but anything you want more perm do in the servers section.
tbh I not sure if /where the monitoring and management setting are stored.
Aug 14, 2019 07:00 AM|brutlex|LINK
Hi and thanks for your input.
The recycling of the application pool was indeed the reason it resets the Monitoring and Management states. I am still trying to find a clean solution for this but at the moment the best I could do is check for the current state through an external program
I wrote which has a state defined in which the server should be and checks in intervals if it matches, if not, then changes it.
Second thing is that Management and Monitoring settings are not stored in shared as those are composed of dynamic states and counters which can be unique for each node. I have used Microsoft.Web.Administration namespace in my .NET Framework program to read
these out and then calculate the sum of all nodes in order to get global counters for the entire ARR cluster, as well as execute methods described in arr_schema.xml.
Hope this gives at least a bit more insight to someone in the future.