Forum Discussion
5 Replies
- hooleylistCirrostratus
Hi David,
Take a look here for some tips:
SOL10430 - Causes of uneven traffic distribution across BIG-IP pool members https://support.f5.com/kb/en-us/solutions/public/10000/400/sol10430.html
Aaron
- David_KrainesssNimbostratus
Thanks, I reviewed that link, but it does not help my specific issue. I am getting fairly even traffic distribution, We have 1000s of clients/theads and teh stats are fairly even.
My question is a given client thread seems to stick to a member server for a given period of time. Why does it not hop around?
- BinaryCanary_19Historic F5 AccountThink HTTP Keepalives.
- Kevin_StewartEmployee
HTTP is stateless, but TCP really isn't, and that's what persistence gets stuck to in the absence of any other mechanisms. Depending on how soon you're doing it, hitting the refresh button in the browser will probably all happen within the same TCP session. Do you have keep-alives enabled? Is this HTTP 1.1 traffic?
- David_KrainesssNimbostratus
OK, thanks for the advice.
Apparently the close connection at the application layer is not actually closing the socket. We are using a Java client to Jetty.
When we added a connection close http header record, we now see each request hop around.
Thanks again!