As several other members of the list have pointed out, to truly ‘see’ if a 
node, either web or app is ‘on’….you need to perform some exercise of that 
node.  In my case, this involved writing some Java code that does the 
exercising…on the web it took the form of a servlet that logs into the MidTier 
in question and verifies it gets to the home page properly.  This servlet is 
monitored directly by the Web load balancer and if the check doesn’t return 
properly the web node is taken out of the rotation.  Secondly, I wrote a little 
java app that logs into the app server directly (check 1, can it log in), and 
performs a query against the user form (check 2, can I do a query), and then 
makes a service call to check and see if the proper results are returned (check 
3, is it responding to other types of calls)….if any of these 3 checks doesn’t 
work an alarm is generated….this app check is unfortunately not tied directly 
into the load balancer…but because monitoring performs this check routinely…we 
can take corrective actions relatively quickly if one of the app nodes goes 
offline….all of this was necessary because a simple ‘is the port active’ or ‘is 
the web server online’ wasn’t sufficient for us to determine that the processes 
were functional.

 

From: Action Request System discussion list(ARSList) 
[mailto:[email protected]] On Behalf Of patchsk
Sent: Sunday, October 30, 2011 11:18 PM
To: [email protected]
Subject: load balancer ars health check

 

** Just curious to see what are the different health check mechanisms being 
used for ARS in a load balanced env in different companies.

We are using F5 and we have load balancing for Midtiers and also for ARS.

For midtiers our health check was to monitor http://webserver:port/arsys/home 
url. And no issues with this.

For ARS we are using tcp monitor on the ARS tcp port.

However recently we are seeing on ars side that the ars tcp port is responding 
fine but the arserver itself is hung, so the load balancer is detecting as if 
the ARS is up and running but users end up getting timeouts during login.

So how are you avoiding this situation in your organization?

Are you guys writing perl/java/c login scripts and using these with the 
loadbalancer as health checks?

_attend WWRUG12 www.wwrug.com ARSlist: "Where the Answers Are"_


_______________________________________________________________________________
UNSUBSCRIBE or access ARSlist Archives at www.arslist.org
attend wwrug12 www.wwrug12.com ARSList: "Where the Answers Are"

Reply via email to