Forum Discussion
NimbostratusMonitoring a pool
Hi Guys,
Quick question, does anyone know the order of operation when a pool member fails a health check ie. if i have http monitor to do a 5sec poll and a 16sec fail, during that time does the f5 still send traffic to that member if it has failed 2 or more checks but is still in the process of being marked down?
Also if the server is being marked down in that .5 sec and a connection comes in does it get re-sent to another member?
i tried to search f5's docs but they dont really go into how an inital connection setup is handled if a node is in failing state but not marked down.
thanks
7 Replies
giltjrThe F5 will continue to attempt to send work/connections to the pool member until 16 seconds after the 1st failure. So from second 1 through second 15 it will continue to try and send connections to that member.
If you are using some type of persistence, what happens to a connection that was persisted to the now down server will depend on what you have selected for the "Action On Service Down" option on the Pool definition.
tidenz_92110Thanks for the response, i was more after what happens between 15.5-16 secs when the member is in the process of being marked down. I have a requirement where users want 100% availabilty (i know right :)) but there application will send data and not care about a received response. its broken i know
- giltjr
LTM will send requests until the member is marked down. Once LTM says its down, no more requests will be forwarded to that member. Until it is down, requests will be forwarded, but no response will be received so the browser will timeout after awhile and display "host not responding" message.
" ... but there application will send data and not care about a received response. "
I not sure I understand this. If the pool member is not responding, then the application is not responding. So what is the application sending?
- greeblesnort
It sounds like you want to build an action around the LB_FAILED event: https://devcentral.f5.com/wiki/iRules.LB_FAILED.ashx
- tidenz_92110Thanks for the link, it gives me a better idea of how it operates, but it only covers the initial syn sequence will probably need to tcpdump and see if i can model the break during and after the handshake.
greeblesnort_33It sounds like you want to build an action around the LB_FAILED event: https://devcentral.f5.com/wiki/iRules.LB_FAILED.ashx
- tidenz_92110Thanks for the link, it gives me a better idea of how it operates, but it only covers the initial syn sequence will probably need to tcpdump and see if i can model the break during and after the handshake.
