Forum Discussion
Dean_M_21414
Nimbostratus
NimbostratusMonitoring wrong?
Hi all,
Just configured a pool with 2 x Server 2008 members (using port 443). The LTM is managing only 443 (https) traffic from the VS to the Pool ; no ssl offload, no NAT, no SNAT, no http profiles, no priority groups. Just using the device as a 'vanilla' LTM. The load bal method is Least Conn (member), yet the statistics for the pool shows that one member is getting a lot more traffic (approx 35x) than the other. The only addition is a source address persistence of 600s.
As an F5 noob, where do I start looking? Is the monitoring wrong? Where can I check?
Regards
Dean
hoolioCirrostratus
Hi Dean,
Here's a SOL to start with:
SOL10430: Causes of uneven traffic distribution across BIG-IP pool members
http://support.f5.com/kb/en-us/solutions/public/10000/400/sol10430.html
Aaron
Dean_M_21414Hello all,
After gathering stats for abt 20 hours, I suspect that the issue is one of the number of connections. In 20 hours, we've had a total of about 380 connections , split in a 5:1 ratio. Comparing this to our busier Pool, which has has 20 thousand connections over the same period, yields an explanation: a single connection in the quiet pool is completed 5 times out of 6 BEFORE a new connection is initiated. Thus, for 5 out of 6 connections, the 'first' server is given the request because there are no other connections. This is probably borne out by the fact that the busier pool is far more evenly dstributing the load between the two members.
Dean
PS, Aaron, tks for the support doc.- Dean_M_21414...and we're running 9.4.5, so this probably applies:
http://support.f5.com/kb/en-us/solutions/public/7000/900/sol7979.html
Regards
Dean - hoolioNice troubleshooting. Thanks for posting what you found. Upgrading to 10.2 should resolve the CR you found and provide a lot more fixes.
Aaron