Full Reconnect Behavior

[scottf]

trafficstars

• [x] Server Pool

A server list pool is maintained. There is a complete implementation in Java for reference.

• [x] Max Retries: How many times each server in the pool should be tried.
• [x] Jitter
• [ ] Tls Jitter
• [ ] Server List Callback

When connecting, server returns list of URLs to other servers in the cluster. Client should store them and use when reconnecting, but users need to be able to disable this option, meaning that client ignores the returned additional URLs and use only the ones passed by a the user.

• [ ] Ignore Discovered Urls:

Clients can ignore discovered servers instead using just bootstrap.

[Jarema]

Already part of the released alpha client.

[scottf]

There is a full specification for reconnect behavior. In my head right now, but partially started in this ADR https://github.com/nats-io/nats-architecture-and-design/pull/202. I will work on it.

@mtmk Please see me in the meantime to discuss.

[mtmk]

This is mostly implemented. We will revisit.

[mtmk]

@scottf, I Updated the todo above. Let me know if I missed/messed something.