mirror of
https://github.com/KumoCorp/kumomta.git
synced 2026-10-26 08:00:35 +00:00
Wez Furlong
db5a6dd1ce
A couple of scenarios we're shooting for here: * Startup ordering races, or other "burps" due to eg: restarting the tsa daemon to update its config * For "replication" scenarios, you may be running multiple instances of tsa-daemon on multiple nodes, and list them all in both the publish and subscribe lists on the clients. In that situation the client will try to read from each node; the data its reads should be essentially the same from each of them, and it is fine if only one out of the set responds, or honestly, even if none of them respond in the moment, as we'll eventually be able to read that data. Errors making those http requests are logged to the diagnostic log, but the shaping data is otherwise allowed to load.