core: implement network ID detection check on startup

[hrharder]

Functional but a work in progress, a few outstanding points (see notes section).

Overview

Closes #237

• Add a Metadata collection to MeshDB to store the networkID we are starting up with
• Add a check during startup that checks if a networkID has been set, compares to the one loaded from the environment, and exits if they do not match.
• The detection check (function initNetworkId) also stores the network ID during the initial startup (no previous DB)

Steps for testing:

1. Start mesh on network N (let it run, collect orders, etc)

P2P_LISTEN_PORT=<port> ETHEREUM_RPC_URL=<url> ETHEREUM_NETWORK_ID=3 ./mesh

2. Kill the process, and restart with a new network ID:

P2P_LISTEN_PORT=<port> ETHEREUM_RPC_URL=<url> ETHEREUM_NETWORK_ID=4 ./mesh

You should see:

{"level":"error","msg":"Mesh previously started on different Ethereum network; switch networks or remove DB","time":"2019-07-23T16:38:09-07:00"}
{"error_string":"expected networkID to be '4', got '3","level":"fatal","msg":"could not initialize app","time":"2019-07-23T16:38:09-07:00"}

Notes

Still a WIP due to me having questions about the following:

• Is there a preferred key for the new collection other than []byte("metadata")? (see here)
• Are there any changes to method names, comments, etc you would like to see?
• Am I handling the mismatched ID case as you would expect? (see here)

[albrow]

@hrharder it looks like we're making the problem harder than it needs to be. I made some suggestions to help simplify the approach.

[albrow]

For testing, we can just call initNetworkID directly and make sure that it behaves as expected. For example, calling initNetworkID twice with two different network IDs should result in an error.

[hrharder]

@albrow I believe I addressed all your comments, please advise if I missed something. Will plan on pushing a test soon.

[albrow]

Just requested a few more changes. Looks pretty good!

[hrharder]

The commit I just pushed should address all the comments. My only remaining questions are:

• Should SaveMetadata accept the networkID or a meshdb.Metadata object?
• Right now, GetMetadata returns any error coming from FindByID, is that OK? or should I explicitly return a ErrNotFound?

[hrharder]

@fabioberger do you have any preference on the signature of the SaveMetadata method? IE should it accept a Metadata value, or just the network ID int? I'm unsure which would be more extensible in the future with more possible metadata values.

[albrow]

Looks great. Thanks @hrharder!