Too many duplicated tcp connections from aggregated apiserver to one member cluster

[lcw2]

What happened:

As the requests to access the aggregated API server increase, the number of connections will grow more, sometimes, the number of connection is 3000

What you expected to happen:
a hostIP should have a tcp connection

How to reproduce it (as minimally and precisely as possible):
Execute tosfedctl get pods requests concurrently, then monitor the number of connections.

Anything else we need to know?:
I think the root case is,

every reqeust will create a httptransport, so can't reuse a TCP connection.

Environment:

• Karmada version: 0.9
• kubectl-karmada or karmadactl version (the result of kubectl-karmada version or karmadactl version):
• Others:

[XiShanYongYe-Chang]

Hi @lcw2, thanks for your feedback. Would you like to try to modify the logic to reuse TCP connections?

In addition, can you provide your detailed test procedure, I want to test it too.

[lcw2]

run this script:
#!/bin/bash while true do karmadactl --kubeconfig xxxxxxx get pods sleep 0.1 done

1. docker ps | grep aggregated-apiserver
2. docker inspect $containerID | grep Pid
3. nsenter -n -t $Pid
4. netstat -anp | grep <IP of member's cluster >

[lcw2]

/assign lcw2

[XiShanYongYe-Chang]

@lcw2 Thanks a lot~

[XiShanYongYe-Chang]

/kind feature

[yanfeng1992]

ping @lcw2 Any progress?

[lcw2]

@yanfeng1992 I'm coding , wait some days.