Comment by deathanatos
1 hour ago
> Restarting all the servers would result in many clients reconnecting several times. It was better to avoid it when possible.
As a sibling says, you need a "reconnect now" in the protocol. (GOAWAY, in HTTP.)
In addition to what the sibling says, if you have some sort of cordoning/graceful drain facility at the traffic level, you can also prevent the "several times" bit: bring new, patched nodes online. Disallow new connections to the outgoing nodes. Drain the outgoing nodes. Decommission them.
(I.e., only permit reconnects to patched nodes.)
No comments yet
Contribute on Hacker News ↗