Skip to content Skip to footer

Help Center

< All Topics
Print

ADR 0016: Consensus Parameters Change Proposal

Component

Oasis Core

Changelog

  • 2022-09-15: Initial version

Status

Proposed

Context

Currently consensus parameters can only be changed with an upgrade governance proposal which is effective but not very efficient. Upgrades require downtime during which binaries need to be updated, nodes restarted and synced, consensus network version has to be increased etc. We would like to avoid this cumbersome procedure and change the parameters of a consensus module as fast and as simple as possible without affecting the performance of the consensus layer.

Decision

Implement governance proposal which changes consensus parameters only.

Implementation

New proposal

A new type of governance proposal named ChangeParametersProposal should be added to the consensus layer. The proposal should contain two non-empty fields:

  • the name of the consensus Module the changes should be applied to, and,
  • a CBOR-encoded document Changes describing parameter changes.
// ChangeParametersProposal is a consensus parameters change proposal.
type ChangeParametersProposal struct {
// Module identifies the consensus backend module to which changes should be
// applied.
Module string `json:"module"`
// Changes are consensus parameter changes that should be applied to
// the module.
Changes cbor.RawMessage `json:"changes"`
}

Both fields should be validated before proposal submission to avoid having invalid proposals with empty fields. A more in-depth validation should be done by consensus modules during submission to ensure that the encoded Changes are complete and well-formed and that there is exactly one module to which changes will be applied.

// ValidateBasic performs a basic validation on the change parameters proposal.
func (p *ChangeParametersProposal) ValidateBasic() error {
// Validate that both fields are set.
}

The new proposal should be added to the ProposalContent. The extension should still allow only one proposal at a time, so we must not forget to update the code responsible for validation.

type ProposalContent struct {
...
ChangeParameters *ChangeParametersProposal `json:"change_parameters,omitempty"`
}

Parameter changes

Each consensus module should carefully scope which parameters are allowed to be changed. For example, a governance module could allow changing only the gas costs and the voting period, while the staking module would allow changing all parameters.

// ConsensusParameterChanges define allowed governance consensus parameter
// changes.
type ConsensusParameterChanges struct {
// GasCosts are the new gas costs.
GasCosts *transaction.Costs `json:"gas_costs,omitempty"`
// VotingPeriod is the new voting period.
VotingPeriod *beacon.EpochTime `json:"voting_period,omitempty"`
}

To prevent invalid proposals being submitted, ConsensusParameterChanges should expose validation method which can be used to check if changes are valid (e.g. changes are not empty, parameters have the right ranges).

// SanityCheck performs a sanity check on the consensus parameters changes.
func (c *ConsensusParameterChanges) SanityCheck() error {
// Validate changes.
}

How changes are executed is up to the module implementation.

// Apply applies changes to the given consensus parameters.
func (c *ConsensusParameterChanges) Apply(params *ConsensusParameters) error {
// Apply changes.
}

Submission

When a new ChangeParametersProposal is submitted a basic validation is performed first which checks whether the Module name and Changes are set correctly. Afterwards, a validation message is broadcasted to all modules requesting them to validate the proposal. Only the module for which Changes are intended should act and reply to the message, other modules should silently ignore it. In case no module replies, the proposal is immediately rejected as not being supported.

The module should carefully examine the proposal, check whether the proposal is well-formed, Changes are not empty and deserialize correctly to the expected format, deserialized parameter changes are valid etc. If all checks succeed, the module should respond with a confirmation message. Otherwise, an error describing why proposal is invalid should be returned as a response.

Note: Validation at this stage cannot always be complete as valid parameter values are not necessary independent of each other. If multiple proposals are being executed at the same time, the resulting parameters can be invalid even though validation of each proposal passed. Therefore, another validation is required when the proposal is about to be executed.

Execution

If ChangeParametersProposal closes as accepted (vote passed), the governance module will execute the proposal by broadcasting a message containing the proposal to all modules. Notification can be done using the same message dispatch mechanism as in the submission phase. Once messages are delivered, only one module will act and try to apply Changes.

That module should first fetch current consensus parameters, then apply proposed Changes and finally validate the result. Validation of parameters is necessary as mentioned in the submission phase. If validation succeeds, the consensus parameters are updated and proposal is marked as passed. Otherwise, the proposal is marked as failed and the proposed parameter changes are discarded.

// SanityCheck performs a sanity check on the consensus parameters.
func (p *ConsensusParameters) SanityCheck() error {
// Validate parameters.
}

How to enable the new proposal

Adding a new proposal type is a consensus breaking change. To make it non-breaking we introduce a new governance consensus parameter which disables the new type by default and can be enabled via governance. When disabled, the governance module will treat the new proposal type as invalid, thus not violating the consensus.

type ConsensusParameters struct {
...
// EnableChangeParametersProposal is true iff change parameters proposals are
// allowed.
EnableChangeParametersProposal bool `json:"enable_change_parameters_proposal,omitempty"`
}

Consequences

Positive

  • Agile and zero-downtime consensus parameter changes.
  • Separation of consensus parameter changes and consensus upgrades.

Negative

  • Introduction of a new governance consensus parameter which enables new proposals in the upgrade handler. New parameters can always be considered as a minor disadvantage as they usually increase the complexity of the code.

Neutral

References

No references.

Table of Contents