Zenoh fails to reconnect when running with lowlatency and without scouting/gossip

[esemeniuc]

Describe the bug

I'm running pub/sub example, meant for servers running in production. Multicast and gossip expose more risk, so they are disabled. I'm have trouble reconnecting after the publisher restarts when in low_latency mode.

To reproduce

Please see this diff: https://github.com/eclipse-zenoh/zenoh/compare/0.11.0-rc.2...jito-labs:zenoh:eric/bug-repro?expand=1

1. Start Publisher on machine 1
2. Start Subscriber on machine 2, let it receive messages
3. Restart Publisher on machine 1
4. Subscriber never reconnects

System info

Arch, linux 6.6.28-1-lts
Rust version: using default in toolchain file

[YuanYuYuan]

Hi @esemeniuc! Thanks for reporting the bug. I can confirm the issue doesn't occur with the default universal transport. I'll investigate the reason.

[esemeniuc]

I noticed that the publisher can use low latency mode without issues. The subscriber is the one bugging out.

[YuanYuYuan]

I noticed that the publisher can use low latency mode without issues. The subscriber is the one bugging out.

Yes, it is. We didn't handle the link closing properly. #1030 should fix this issue.