Forum Discussion

Aoi_5678's avatar
Aoi_5678
Icon for Nimbostratus rankNimbostratus
Feb 23, 2012

Bigip fastL4 Protocol

I have the following settings in the bigip and i figured out comparing another bigip using TCP protocol that the active connections showing in the performance data graphs became about 4 times more than the real connections taken from the bigip command line.

 

 

 

Also when i changed the Idle Timeout time from 300(default) to 60 the active connections showing in the performance data graphs became about 2 times more than the real connections taken from the bigip command line.

 

 

 

Protocol:fastL4

 

 

Settings

 

 

Idle Timeout:300

 

 

 

I think that the connections dependicity is on the server and Idle Timeout in the BigIp settings on the above example can set it to 0 without any problem.

 

 

 

Can you confirm if this is correct?

 

3 Replies

  • i do not think it is good to set immediate idle timeout for a connection which needs to maintain state.

     

     

    sol7606: Overview of BIG-IP idle session timeouts

     

    http://support.f5.com/kb/en-us/solutions/public/7000/600/sol7606.html
  • Hamish's avatar
    Hamish
    Icon for Cirrocumulus rankCirrocumulus
    Remember that the fastL4 profile will also by default send an RST to both ends of the connection when it expires. So if you set it too low, you will actually kill off valid connections. Which may not always be what you want (I often set low timeouts on fastL4 profiles attached to network virtual servers. But also disable RST and enable loose-open/loose-close which makes the VS more like a router.
  • thank you for your response.

     

     

    There will become many troubles in the system if the current valid connections killed.

     

    So as you explained to me it is better not to set idle timeout too low.

     

    If i will change the existing idle timeout settings i will consider the above settings to disable the RST etc. but fist testing it to a testing environment.