Source balance Impairment

I’ve switched from roundrobin to source balance,
but I got a very unbalanced allocation, I also tried to drain the most active backend but redistribution is odd.
and I’ve traced more than 170 distinct client IP

Do you have 1 user/browser per source IP? Or do you have companies with many users behind a single IP address?

A hash of the source IP is a simple balancing algorithm on one hand, but has a lot of problems like this in real production workloads. You really should try to avoid it.

1 Like

Thanks @lukastribus , I will try this alternative on test farm

balance leastconn
stick-table type ip size 20k expire 4h
stick on src

I used this command:
netstat -ant | fgrep ':443 ’ | fgrep ESTAB | awk ‘// { print $5 }’ | cut -d: -f1 | uniq | wc -l
for isolate uniq IP on https

This is the picture after 2 days, source ip is maintained on same server and balanced
:grinning: :gem:

1 Like