Suppress bootstrap error

[zonque]

When used in private network mode with Bootstrap set to [], the daemon currently spits out the following message ever 30 seconds:

ERROR       core: no bootstrap nodes configured: go-ipfs may have difficulty connecting to the network bootstrap.go:144

While it is good to mention, I see two issues with this:

• In regular environments, running the daemon with no bootstrap nodes will in fact render it unusable, so there should be an error at daemon start time instead of a repetitive message
• For private network setups where nodes discover themselves via MDNS, this message is unnecessary.

This PR changes both the above, with one patch for each of them.

[zonque]

Hmm, so some test now fail because they instantiate daemons without bootstrap nodes. Can anyone give me a hint on how to quickly fix that?

[Stebalien]

For private network setups where nodes discover themselves via MDNS, this message is unnecessary.

I'm not sure I agree with this. Really, MDNS and private networks are unrelated (unless you mean private networks via private VPNs). Most private networks (that use swarm keys) will want private bootstrap nodes.

However, I do agree that this is way too noisy. It may make sense to just print a warning once and never again.

[Stebalien]

The tests are probably failing because bootstrappers can be added and removed at runtime.

[zonque]

I'm not sure I agree with this. Really, MDNS and private networks are unrelated (unless you mean private networks via private VPNs). Most private networks (that use swarm keys) will want private bootstrap nodes.

However, I do agree that this is way too noisy. It may make sense to just print a warning once and never again.

Hmm, okay. Agreed.

I've force-pushed a new version that does just that.

Thanks for your input!

[Stebalien]

LGTM but I'd like someone else to take a quick look for a sanity check.

[schomatis]

LGTM, I would change the error for a warning (but it's just a personal preference so feel free to ignore it).

[schomatis]

The tests are probably failing because bootstrappers can be added and removed at runtime.

Can this be merged with the test errors?

[Stebalien]

LGTM, I would change the error for a warning (but it's just a personal preference so feel free to ignore it).

We should, but I'd like to enable warnings by default first. Otherwise, users will never see it.

[Stebalien]

Can this be merged with the test errors?

That has been fixed. The first iteration aborted the construction of the periodic bootstrapper when no bootstrappers were configured. This one just warns and continues.