What am I missing? Please let me know if there is other information I can supply. Thank you.
Next time tag your question with parity, to make sure I wont miss it :p
Cross-client private networks: The problem
Geth is a client written for Ethereum. It was one of the first official reference implementations and was never meant to run anything else but Ethereum. If you want to configure Geth for any other chain configuration, you would have to fork the source-code and create your own client implementation based on your desired rule-set. This happened for instance for Expanse (Gexp) and Ethereum Classic (Getc).
Parity, however, was created much later than Ethereum itself, by a team that was initially involved with the C++ Ethereum client (Eth). After they (Gavin Wood, et al.) founded Ethcore (now Parity), they created a client with a much broader vision, a client that is not supposed to only run Ethereum.
Parity allows more abstraction in the core and therefore, all you need to start a new chain is a so called Chain Specification file which describes the whole chain parameters, including consensus rules, validator engines, and so on. The most visible consequence is that Ethereum, Ethereum Classic, and Expanse (just to pick up the examples form above) do not need to maintain their own copy of the Parity source-code in order to support their project and their consensus rules. Parity just works out of the box with the --chain
parameters foundation
, classic
, or expanse
.
In addition, the --chain
parameter allows configuring an entirely custom chain with pluggable consenus, custom transitions, and the validator engine you prefer.
In contrast, Geth only allows to specify a custom Genesis which is only one piece of the puzzle for a complete chain specification. And therefore you will have trouble to connect a Geth node to a custom network of Parity nodes. However, you basically have two options to run a custom cross-client network with both Geth and Parity nodes, explained below.
Best-practice: "Geth First"
Now, since you have fewer chain configuration options in Geth than in Parity, I'll call the most obvious idea "Geth First". You chose a network ID and create a custom Genesis, and initialize your new network with geth init
. You can fire up a miner thread and a couple more Geth nodes at this point to verify the network is exactly what you need.
Now, you can start integrating Parity nodes. There is a tiny rust tool keorn/parity-spec that converts your custom Geth genesis file into a full Parity chain specification file.
cargo run -- geth-genesis.json
You can pass the output parity-spec.json
directly to you Parity node with the --chain
parameter. For discovery, you can use Geth's admin.addPeer()
commands or Parity's --reserved-peers
functionality.
However, both Geth and Parity are actively developed code-bases and the converter tool is prone to bugs and often users report generated chain-specification files are still leading to consensus issues. If this happens to you frequently, check out the other option below.
Best-practice: "Ropsten Revived - Revived"
The probably most stable and most obvious option to run a private network compatible with Geth, Parity, and probably all other clients out there, is something, I would call "Ropsten Revived - Revived".
Just take known, working network that is supported by all clients, like the Ropsten public testnet, and slightly modify it. To ease this process, I created presets for both the Genesis and the Parity Chain Specification at 5chdn/crossclient-chainsepc. It contains a geth.json
and a parity.json
mimicking a Ropsten revival with network ID 1337 (0x539
).
The downside of this is probably the unsatisfied desire to add more complex modifications to your custom private development network. However, the predefined genesis and chain-spec offer a good starting point to get you bootstrapped with whatever network you have in mind.
Best Answer
The --nodiscover option works the other way. If you run geth with this option, the geth node will not search for any peer and not connect to any peer. So yes, you will need a bootnode list or add peers manually if you use the --nodiscover option. However, you need it for your node, not for the others.
As far as I know, you cannot hold back other nodes from connecting to your node, besides changing or blocking the port used for peer discovery. The default for that port is 30303, that's right, you can change it with the --port option.
To answer your question: --nodiscover is not redundant behind a firewall without port forwarding, since it achieves a different goal.