ts_runtime: drop unregistered map request update log to debug - #373
Conversation
Signed-off-by: Dylan Bargatze <dylan@tailscale.com>
nrc
left a comment
There was a problem hiding this comment.
If this is still an error any time except the first, is it possible to log that error? Because otherwise it looks like it's just silently eaten. Don't worry if that is a pain though. In any case, I think a comment here (about when this is/isn't an error) would be useful
There was a problem hiding this comment.
wait, how does this codepath even occur? oh, i see, it must be that it's getting new endpoints from netmon, which then triggers an attempt to tell control about them. @nrc yeah this could happen repeatedly, if we're not registered but keep getting updates from other actors that trigger an update request.
this indicates actually a more substantial bug, which i think we're only saved by from the fact that the derp latency measurement has to wait until the first netmap to see the derp map, and it succeeding necessarily triggers an update_map_request. if we didn't have that feedback path, the endpoints could go unobserved. i'll make a follow-up pr -- imo we shouldn't wait for it for 0.5.0 though, since it can't actually trigger
Drops a spurious log message to DEBUG-level; with the current architecture, this is expected to happen once on startup.