Forum Discussion
GIS1STOP_182395
Nimbostratus
NimbostratusDoes LTM really support Zero downtime?
In web application system, there are two IIS web servers running the same web site. Above them, there is one LTM as load balancer. I was told, if one web server is shutdown, it will take up to 90 seconds for LTM knowing the one web server is shutdown and then redirect future requests. So during that 90 seconds, some requests to our web site will fail.
According to LTM product page: https://f5.com/products/modules/local-traffic-manager. It could be zero down time. How can I tell my LTM administrator to configure LTM so that when one web server is failed, all request will immediately redirect to the another running web server?
Thanks
Hi, the monitor interval and timeout settings are user configurable. You could set the interval for one second and the timeout for 4 seconds, after which a connection load balanced to that downed pool member would be handled according to the "Action on Service Down" setting in the pool configuration.
Too aggressive interval/timeout settings will probably put too much traffic on the wire and too much load on your servers...you'll want to adjust them appropriately for your application.
If you are using the IIS iApp, the template does default to 30/91 for interval/timeout, but allows you to specify the monitor interval and adjusts the timeout to (interval x3) + 1 second. If you use a 5 second interval, the timeout will be 16 seconds.
Hi, the monitor interval and timeout settings are user configurable. You could set the interval for one second and the timeout for 4 seconds, after which a connection load balanced to that downed pool member would be handled according to the "Action on Service Down" setting in the pool configuration.
Too aggressive interval/timeout settings will probably put too much traffic on the wire and too much load on your servers...you'll want to adjust them appropriately for your application.
If you are using the IIS iApp, the template does default to 30/91 for interval/timeout, but allows you to specify the monitor interval and adjusts the timeout to (interval x3) + 1 second. If you use a 5 second interval, the timeout will be 16 seconds.
GIS1STOP_182395Mike, Thank you very much.
- GIS1STOP_182395
Mike, I do agree with you that I could have LTM administrator to adjust intervals. However, what does "zero downtime" mean in LTM product home page https://f5.com/products/modules/local-traffic-manager? I feel it is misleading to me. isn't it?
- nathe
Cirrocumulus
Other ways that the LTM can improve application availability, and get somewhere to your zero downtime goal is Passive Monitoring (https://support.f5.com/kb/en-us/solutions/public/7000/400/sol7440.html) and Action On Service Down (https://support.f5.com/kb/en-us/solutions/public/15000/000/sol15095.html).
I know your looking at the bigger picture but i hope this is useful anyway.
N
JGCumulonimbus
On caveat is that if you configure you monitor to run too frequently, you might run into an issue of port exhaustion, or a port re-use issue. You might need to match / tune the network stack of your app server for this to work.