I'm working on one the issue wherein I have been taked to check for a better healtch monitor of the servers from LTM.
Issue scenario: we are using http based monitor to check the health of the servers. CPU utilization of one of the member server was high and it was causing some application performance issue. However on LTM the server was keep responding within the faildetect interval so the status of the server was not down on LTM.
Is there a better to tweak the health check of the application from LTM based on the performance of the application servers?
The easiest options would be to select ‘Fastest’ load balancing method.
The theory is servers that are more overloaded will have a slower response time, so a smaller ratio of connections will go to these servers.
If you want to perform an advance health check you would need to discuss this with your application team, with them perhaps writing a dedicated web method that F5 could check performance
I once worked at a place that wrote a method that allowed an API call trigger a DB query, a successful response proved the full application stack was working.
You could use a similar approach to only return a successful response if the server is within certain performance thresholds such as having less than 50% CPU utilisation.
Another alternative would be to use an external monitor to query the pool member via SNMP to return performance metrics.