[lvs-users] estimation_timer take too much cpu usage.

唐阳-技术中台 tangyang at zhihu.com
Thu Jul 9 04:57:09 BST 2020

On my kubernetes node , there are some tcp packets randomly delayed between
two containers on one node.
I notice that estimation_timer softirq execute about 30ms+.(traced by this
tool: https://github.com/bytedance/trace-noschedule)

there are about 3000+ local address shown by 'ipvsadm -l'.

Hyper threading is enabled and there are 72 logical cores on my machine.

Does anyone know how to reduce the execution time of estimation_timer

Thank you.

More information about the lvs-users mailing list