white-park-86714
10/07/2022, 2:15 PMnetstat -t | wc -l
(TIME_WAIT conns) (on fpm -> to redis)
is showing values between 2k and 4k while sysctl net.core.somaxconn
shows a value of 4096
On our old cluster anything is fine and works well while only 100 TIME_WAIT connections for the kick from net.ipv4.tcp_fin_timeout=60
I allready tried to lower fin_timeout to try to proove that the waiting connections are the issue without any success.
Rancher is removing
securityContext:
sysctls:
- name: net.ipv4.tcp_fin_timeout
value: "30"
from our yaml.