cmd/containerboot: switch to IPN bus monitoring instead of polling. #6658
Conversation
|
CIFuzz failure is some runtime error in the fuzz infrastructure, not in the change. |
| case ipn.NeedsMachineAuth: | ||
| log.Printf("machine authorization required, please visit the admin panel") | ||
| case ipn.Running: | ||
| // Technically, all we want is to keep monitoring the bus for |
There was a problem hiding this comment.
Would it instead make sense to run this in a goroutine and have that push the relevant notifications to the main goroutine? That way you can keep the cancel the watcher goroutine if something fails or times out?
There was a problem hiding this comment.
Possibly. I was trying to avoid channels and goroutines, because it's already quite tricky to sequence things when it's just ~1 linear state, and as long as the IPN watch is active you have to consume the updates without blocking. Splitting the logic would force me to reimplement a chunk of ipn/ipnlocal to coalesce changes and track overall state.
I'll give it a try in a followup change, getting this one to pass tests was tricky enough that I don't want to lose it.
We still have to shell out to `tailscale up` because the container image's API includes "arbitrary flags to tailscale up", unfortunately. But this should still speed up startup a little, and also enables k8s-bound containers to update their device information as new netmap updates come in. Fixes #6657 Signed-off-by: David Anderson <danderson@tailscale.com>
danderson
force-pushed
the
danderson/containerboot-localapi
branch
from
5de0962
to
8a675dd
Compare
We still have to shell out to
tailscale upbecause the container image's API includes "arbitrary flags to tailscale up", unfortunately. But this should still speed up startup a little, and also enables k8s-bound containers to update their device information as new netmap updates come in.Fixes #6657
Signed-off-by: David Anderson danderson@tailscale.com