Forum Discussion

Verbal_92765's avatar
Verbal_92765
Icon for Nimbostratus rankNimbostratus
Mar 12, 2012

health monitor to remove nodes from pool with no drian

Hello All, We are running an F5 big IP 3600 series in front of a Windows web farm. We are using automated deployments to each web node. In order to remove web nodes from the pool when deploying, I have setup health monitor that checks for the content of an .htm file on each node. Send String GET /maintenance.htm\n\n Receive String good Receive Disable String maintenance This works well apart from active connections are still allowed, this means files are always in use etc. I do not wish to wait for connections to drain or expire when this health monitor kicks in. I would simply like to remove all connections and disable the node. Any ideas how I can change this health monitor to do just this? Kind Regards
  • have you tried "action on service down"?

     

     

    LTM: Action on Service Down by Deb

     

    https://devcentral.f5.com/Tutorials/TechTips/tabid/63/articleType/ArticleView/articleId/179/LTM-Action-on-Service-Down.aspx
  • This is set to 'Reselect' at the moment, but still active connections are coming to the node that has been marked down by the monitor.

     

  • This is set to 'Reselect' at the moment, but still active connections are coming to the node that has been marked down by the monitor.how did you notice it? have you done tcpdump on bigip?

     

     

    anyway, does "reject" make any difference?
  • the IIS logs on the web server were still showing active connections.

     

    Will try reject
  • reject should sort all client initiated connections.

     

    Just to double check, the remaining ones are not server initiated connections?

     

    Seen this before, but mostly with link controller customers.

     

    regards,

     

    Christian

     

     

  • for example. If i stop the node on the command line with the command 'bigip node 192.168.0.x down' then there are no new connections.
  • Do you have other pools with that IP:port combination? Disabling the node would mark the server down in all pools, whereas marking a pool member down only affects that pool.

     

     

    If you want the node marked down if a specific service fails, you could create a new monitor with the alias port set to 80 and the action on service down set to reject. Add that monitor to the node address (not the pool) and retest.

     

     

    Aaron