SN doesn't reconnect to RPC

[roman-khimov]

Expected Behavior

If anything bad happens to RPC connection SN should reconnect.

Current Behavior

Endless

error policer/check.go:86 could not get container {"component": "Object Policer", "cid": "cWaXnFSPr5teUdtUu8MFwsm2LJ3W7NanzBpWWmCCmeF", "error": "could not perform test invocation (get): failed to write JSON request (invokefunction / 4): write tcp 159.223.36.204:55904->116.203.249.217

seen on one of testnet nodes.

Possible Solution

Not known.

Steps to Reproduce (for bugs)

Not known, unfortunately. Usually this doesn't happen.

Context

Testnet node number 3.

Your Environment

• Version used: 0.40.1
• Server setup and configuration: T5 testnet
• Operating System and version (uname -a): Debian 11

[roman-khimov]

RPC service itself is available in this case.

[carpawell]

Testnet node https://github.com/nspcc-dev/neofs-node/issues/3.

Bad link? Just 3rd testnet node?

write tcp 159.223.36.204:55904->116.203.249.217

@roman-khimov, are you sure that is not neo-go issue? Was there any lock? Seems like the node did not switch cause neo-go client did not say it had to. Anyway I do not expect neo-go to act like that: it tries to send smth to a missing connection? E.g. on our side, we catch such problems.

[roman-khimov]

Just 3rd testnet node?

Yeah, third one, not intended to be a link.

are you sure that is not neo-go issue?

I don't know. But there is an error returned to NeoFS in this case. And connection should be closed.