Lot's of sockets in CLOSE-WAIT

[ashangit]

On few of our servers during roll restart of forwarder service due to release bump we can observe bad sockets behaviour with lots of CLOSE-WAIT sockets from container to forwarder

> ss -np|grep 33000|grep CLOSE-WAIT|wc -l
26358

Netid  State      Recv-Q Send-Q Local Address:Port           Peer Address:Port      
tcp    CLOSE-WAIT 0      0      127.0.0.1:33068              127.0.0.1:33000               users:(("java",pid=46695,fd=4494))
tcp    CLOSE-WAIT 0      0      127.0.0.1:35035              127.0.0.1:33000               users:(("java",pid=46695,fd=26321))

[ashangit]

If a failure happen when agent is sending data to the forwarder we are will closing the socket but when we do handshake between agent and forwarder if there is a version check issue we don'(t close the connection from the client side.
We'll update the code to do so

[Willymontaz]

Reopened until no more processes containing such agent runs on the cluster / gateways

[ashangit]

The patch solved this issue
If we want to follow state of the patch on our infra we should open a request in our internal repo.
Here we should only report feature, bug... and close when code is merged

[Willymontaz]

Yes, right