Deep dive into Substrate consensus - part 2
We introduced why consensus is important to decentralized computing and storage in part-1 of the deep dive into substrate consensus series.
This guide continues where part-1 ends. Here we will develop an understanding of blockchain finality, and have a grasp of relevant Substrate components involved in block finality. We will also look at why blockchain blocks must be finalized, and highlight common approaches for block finality.
Help us measure our progress and improve Substrate in Bits content by filling out our living feedback form. Thank you!
Review of Block processing
To provide context to the subsequent discussion, let us review how a block is added to the blockchain. The steps below provide a summary of how a block is created and added to the blockchain.
-
Transactions are submitted to a node.
-
Submitted transactions are validated and added to the transaction pool.
-
Valid transactions are ordered based on Substrate's transaction priority system.
-
Valid transactions are used to composed into a block, and a block header is created.
-
The composed block gets executed by the runtime executor.
-
Block is gossiped among peers and validated using the block header.
-
An authorized node may add the executed block to the blockchain.
You can find a detailed description with code examples of how blocks are processed in a previous series here.
Understanding Block Finality
You may have noticed that the steps above did not highlight how one can ascertain the guarantee of the state of the blockchain at a given point in time. We can say that there was no description of how we can know a block is final and its content as well as its order in the blockchain will not change.
Block finality is the second face of a blockchain consensus mechanism. (Block production being the first face of blockchain consensus). It solves the problem of how can peers within a decentralized network know the correct chain to build upon. By extension, it facilitates data consistency and enables external data requests to produce the same result.
There are two widely used approaches to block finality; probabilistic vs provable finality.
Probabilistic finality means that under some assumptions about the network and participants, for example; if a few blocks build on a given block, one can estimate the probability that it is final. Based on these assumptions, at some point in the future, all nodes will agree on the truthfulness of one set of data (ie. block becomes final). This eventual consensus on the finality of a block may take a long time and will not be able to determine how long it will take ahead of time.
Provable finality on the other hand depends on a set of specified rules and processes to agree on the truthfulness of data on the blockchain. Under these specified rules, peers within the network continue to agree on the truthfulness of new and old blocks.
Finality gadgets such as GRANDPA (GHOST-based Recursive ANcestor Deriving Prefix Agreement) or Ethereum's Casper FFG (the Friendly Finality Gadget) are designed to give stronger and quicker guarantees on the finality of blocks - specifically, that they can never be reverted after some process of Byzantine agreements has taken place. The notion of irreversible consensus is known as provable finality.
GRANDPA
GRANDPA design
GRANDPA (GHOST-based Recursive ANcestor Deriving Prefix Agreement) is the finality gadget that is implemented by default for Substrate chains. Finality is obtained by consecutive rounds of voting by validator nodes. Validators execute the GRANDPA finality process in parallel to Block Production as an independent service.
It is important to note that that GRANDPA reaches agreements on chains rather than blocks which greatly speeds up the finalization process, even after long-term network partitioning or other networking failures. This means that as soon as more than 2/3 of validators attest to a chain containing a certain block, all blocks leading up to that one are finalized at once.
The GRANDPA protocol specifies the following to successfully participate in the block-finalization process:
- A GRANDPA voter is represented by a key pair generated using the ed25519 encryption protocol.
- GRANDPA authorities is a set of voters for a given block.
- The GRANDPA authority set id is an incremental counter that is updated under specific conditions.
- A GRANDPA vote is composed of a block hash and the block number.
- A GRANDPA voter can engage in a maximum of 2 sub-rounds of voting. The first sub-round is called the pre-round and the second sub-round is called pre-commit.
- Voting is done by broadcasting voting messages to the network.
- Validators inform their peers about the block finalized in a round by broadcasting a commit message.
- A message is a byte array containing the message to be signed.
- A Vote signature is the signature of a voter for a specific message in a round and is specific for each sub-round.
- A block must have a valid justification. The justification must contain up to one valid vote from each voter and must not contain more than two equivocatory votes from each voter.
- A Voter equivocates if they broadcast two or more valid votes to blocks during one voting sub-round.
To participate coherently in the voting process, a validator must initiate its state and sync it with other active validators. In particular, considering that voting happens in different distinct rounds where each round of voting is assigned a unique sequential round number, a validator node needs to determine and set its round counter equal to the voting round currently undergoing in the network.
The process of joining a new voter set is different from the one of rejoining the current voter set after a network disconnect.
For each round, an honest voter must participate in the voting process by
following Play-Grandpa-Round
. It is important to note that we might not
always succeed in finalizing a block candidate in a round. In this case, another
round of voting is conducted before voting on the next voting round.
At the end of a voting round and block finalization, a Justified Block Header
for a block is appended to the blockchain. This justified block header contains
the following parts; the block header
, justification
, and
authority id
.
To catch up to the latest chain after network disruption, a node (re)joins the network, and requests the history of state transitions in the form of blocks, which it is missing. When a voter node joins the network, it also needs to gather the justification of the rounds it has missed. Through this process, they can safely join the voting process of the current round, on which the voting is taking place.
check here for a full view of GRANDPA specifications.
GRANDPA implementation
From the section above, we can see that GRANDPA is fairly complex. Its
implemntation is contained mainly in three Substrate modules;
sp_consensus_grandpa
, sc_consensus_grandpa
, and pallet-grandpa
.
The sp_consensus_grandpa
and sc_consensus_grandpa
module is part of
external client which facilitates running GRANDPA related consesus related
tasks. They contains all implementation GRANDPA related to peer-to-peer
communication, message and equivocation related tasks. The
sp_consensus_grandpa
module also defines an interface used to interact
with pallet-grandpa
from the external client.
The GRANDPA authority set is implemented like so:
pub struct AuthoritySet<H, N> {
/// The current active authorities.
pub(crate) current_authorities: AuthorityList,
/// The current set id.
pub(crate) set_id: u64,
pub(crate) pending_standard_changes: ForkTree<H, N, PendingChange<H, N>>,
pending_forced_changes: Vec<PendingChange<H, N>>,
pub(crate) authority_set_changes: AuthoritySetChanges<N>,
}
The GRANDPA voter task which carry out voting mechanism is implemented like so:
/// Run a GRANDPA voter as a task. Provide configuration and a link to a
/// block import worker that has already been instantiated with `block_import`.
pub fn run_grandpa_voter<Block: BlockT, BE: 'static, C, N, S, SC, VR>(
grandpa_params: GrandpaParams<Block, C, N, S, SC, VR>,
) -> sp_blockchain::Result<impl Future<Output = ()> + Send>
where
BE: Backend<Block> + 'static,
N: NetworkT<Block> + Sync + 'static,
S: SyncingT<Block> + Sync + 'static,
SC: SelectChain<Block> + 'static,
VR: VotingRule<Block, C> + Clone + 'static,
NumberFor<Block>: BlockNumberOps,
C: ClientForGrandpa<Block, BE> + 'static,
C::Api: GrandpaApi<Block>,
{
let GrandpaParams {
mut config,
link,
network,
sync,
voting_rule,
prometheus_registry,
shared_voter_state,
telemetry,
offchain_tx_pool_factory,
} = grandpa_params;
config.observer_enabled = false;
let LinkHalf {
client,
select_chain,
persistent_data,
voter_commands_rx,
justification_sender,
justification_stream: _,
telemetry: _,
} = link;
let network = NetworkBridge::new(
network,
sync,
config.clone(),
persistent_data.set_state.clone(),
prometheus_registry.as_ref(),
telemetry.clone(),
);
let conf = config.clone();
let telemetry_task =
if let Some(telemetry_on_connect) = telemetry.as_ref().map(|x| x.on_connect_stream()) {
let authorities = persistent_data.authority_set.clone();
let telemetry = telemetry.clone();
let events = telemetry_on_connect.for_each(move |_| {
let current_authorities = authorities.current_authorities();
let set_id = authorities.set_id();
let maybe_authority_id =
local_authority_id(¤t_authorities, conf.keystore.as_ref());
let authorities =
current_authorities.iter().map(|(id, _)| id.to_string()).collect::<Vec<_>>();
let authorities = serde_json::to_string(&authorities).expect(
"authorities is always at least an empty vector; \
elements are always of type string",
);
telemetry!(
telemetry;
CONSENSUS_INFO;
"afg.authority_set";
"authority_id" => maybe_authority_id.map_or("".into(), |s| s.to_string()),
"authority_set_id" => ?set_id,
"authorities" => authorities,
);
future::ready(())
});
future::Either::Left(events)
} else {
future::Either::Right(future::pending())
};
let voter_work = VoterWork::new(
client,
config,
network,
select_chain,
voting_rule,
persistent_data,
voter_commands_rx,
prometheus_registry,
shared_voter_state,
justification_sender,
telemetry,
offchain_tx_pool_factory,
);
let voter_work = voter_work.map(|res| match res {
Ok(()) => error!(
target: LOG_TARGET,
"GRANDPA voter future has concluded naturally, this should be unreachable."
),
Err(e) => error!(target: LOG_TARGET, "GRANDPA voter error: {}", e),
});
// Make sure that `telemetry_task` doesn't accidentally finish and kill grandpa.
let telemetry_task = telemetry_task.then(|_| future::pending::<()>());
Ok(future::select(voter_work, telemetry_task).map(drop))
}
The pallet-grandpa
manages the GRANDPA authority set on the runtime. It
routinely check for changes in the authority set change and enacts them in the
next voting round. The pallet-grandpa
also implements equivocation offence
releted task.
Summary
We were able to get a firm grasp of the concept of blockchain consensus especially how the block authoring works within the framework of Substrate. We appreciated how Substrate implemented out-of-the-box block authoring mechanisms and gained insight into relevant Substrate consensus modules.
We developed an understanding of the following:
- Why blockchains need consensus.
- Functional components of a consensus mechanism.
- Aura implementation.
- BABE implementation.
To learn more about substrate consensus, check out these resources:
Help us measure our progress and improve Substrate in Bits content by filling out our living feedback form. Thank you!