Implementation of the new validator disabling strategy (#2226)

Closes https://github.com/paritytech/polkadot-sdk/issues/1966,
https://github.com/paritytech/polkadot-sdk/issues/1963 and
https://github.com/paritytech/polkadot-sdk/issues/1962.

Disabling strategy specification
[here](https://github.com/paritytech/polkadot-sdk/pull/2955). (Updated
13/02/2024)

Implements:
* validator disabling for a whole era instead of just a session
* no more than 1/3 of the validators in the active set are disabled
Removes:
* `DisableStrategy` enum - now each validator committing an offence is
disabled.
* New era is not forced if too many validators are disabled.

Before this PR not all offenders were disabled. A decision was made
based on [`enum
DisableStrategy`](https://github.com/paritytech/polkadot-sdk/blob/bbb6631641f9adba30c0ee6f4d11023a424dd362/substrate/primitives/staking/src/offence.rs#L54).
Some offenders were disabled for a whole era, some just for a session,
some were not disabled at all.

This PR changes the disabling behaviour. Now a validator committing an
offense is disabled immediately till the end of the current era.

Some implementation notes:
* `OffendingValidators` in pallet session keeps all offenders (this is
not changed). However its type is changed from `Vec<(u32, bool)>` to
`Vec<u32>`. The reason is simple - each offender is getting disabled so
the bool doesn't make sense anymore.
* When a validator is disabled it is first added to
`OffendingValidators` and then to `DisabledValidators`. This is done in
[`add_offending_validator`](https://github.com/paritytech/polkadot-sdk/blob/bbb6631641f9adba30c0ee6f4d11023a424dd362/substrate/frame/staking/src/slashing.rs#L325)
from staking pallet.
* In
[`rotate_session`](https://github.com/paritytech/polkadot-sdk/blob/bdbe98297032e21a553bf191c530690b1d591405/substrate/frame/session/src/lib.rs#L623)
the `end_session` also calls
[`end_era`](https://github.com/paritytech/polkadot-sdk/blob/bbb6631641f9adba30c0ee6f4d11023a424dd362/substrate/frame/staking/src/pallet/impls.rs#L490)
when an era ends. In this case `OffendingValidators` are cleared
**(1)**.
* Then in
[`rotate_session`](https://github.com/paritytech/polkadot-sdk/blob/bdbe98297032e21a553bf191c530690b1d591405/substrate/frame/session/src/lib.rs#L623)
`DisabledValidators` are cleared **(2)**
* And finally (still in `rotate_session`) a call to
[`start_session`](https://github.com/paritytech/polkadot-sdk/blob/bbb6631641f9adba30c0ee6f4d11023a424dd362/substrate/frame/staking/src/pallet/impls.rs#L430)
repopulates the disabled validators **(3)**.
* The reason for this complication is that session pallet knows nothing
abut eras. To overcome this on each new session the disabled list is
repopulated (points 2 and 3). Staking pallet knows when a new era starts
so with point 1 it ensures that the offenders list is cleared.

---------

Co-authored-by: ordian <noreply@reusable.software>
Co-authored-by: ordian <write@reusable.software>
Co-authored-by: Maciej <maciej.zyszkiewicz@parity.io>
Co-authored-by: Gonçalo Pestana <g6pestana@gmail.com>
Co-authored-by: Kian Paimani <5588131+kianenigma@users.noreply.github.com>
Co-authored-by: command-bot <>
Co-authored-by: Ankan <10196091+Ank4n@users.noreply.github.com>
This commit is contained in:
Tsvetomir Dimitrov
2024-04-26 16:28:08 +03:00
committed by GitHub
parent 97f7425338
commit 988e30f102
33 changed files with 775 additions and 700 deletions
@@ -23,7 +23,6 @@ pub(crate) const LOG_TARGET: &str = "tests::e2e-epm";
use frame_support::{assert_err, assert_noop, assert_ok};
use mock::*;
use sp_core::Get;
use sp_npos_elections::{to_supports, StakedAssignment};
use sp_runtime::Perbill;
use crate::mock::RuntimeOrigin;
@@ -127,75 +126,48 @@ fn offchainify_works() {
}
#[test]
/// Replicates the Kusama incident of 8th Dec 2022 and its resolution through the governance
/// Inspired by the Kusama incident of 8th Dec 2022 and its resolution through the governance
/// fallback.
///
/// After enough slashes exceeded the `Staking::OffendingValidatorsThreshold`, the staking pallet
/// set `Forcing::ForceNew`. When a new session starts, staking will start to force a new era and
/// calls <EPM as election_provider>::elect(). If at this point EPM and the staking miners did not
/// have enough time to queue a new solution (snapshot + solution submission), the election request
/// fails. If there is no election fallback mechanism in place, EPM enters in emergency mode.
/// Recovery: Once EPM is in emergency mode, subsequent calls to `elect()` will fail until a new
/// solution is added to EPM's `QueuedSolution` queue. This can be achieved through
/// `Call::set_emergency_election_result` or `Call::governance_fallback` dispatchables. Once a new
/// solution is added to the queue, EPM phase transitions to `Phase::Off` and the election flow
/// restarts. Note that in this test case, the emergency throttling is disabled.
fn enters_emergency_phase_after_forcing_before_elect() {
/// Mass slash of validators shouldn't disable more than 1/3 of them (the byzantine threshold). Also
/// no new era should be forced which could lead to EPM entering emergency mode.
fn mass_slash_doesnt_enter_emergency_phase() {
let epm_builder = EpmExtBuilder::default().disable_emergency_throttling();
let (ext, pool_state, _) = ExtBuilder::default().epm(epm_builder).build_offchainify();
let staking_builder = StakingExtBuilder::default().validator_count(7);
let (mut ext, _, _) = ExtBuilder::default()
.epm(epm_builder)
.staking(staking_builder)
.build_offchainify();
execute_with(ext, || {
log!(
trace,
"current validators (staking): {:?}",
<Runtime as pallet_staking::SessionInterface<AccountId>>::validators()
ext.execute_with(|| {
assert_eq!(pallet_staking::ForceEra::<Runtime>::get(), pallet_staking::Forcing::NotForcing);
let active_set_size_before_slash = Session::validators().len();
// Slash more than 1/3 of the active validators
let mut slashed = slash_half_the_active_set();
let active_set_size_after_slash = Session::validators().len();
// active set should stay the same before and after the slash
assert_eq!(active_set_size_before_slash, active_set_size_after_slash);
// Slashed validators are disabled up to a limit
slashed.truncate(
pallet_staking::UpToLimitDisablingStrategy::<SLASHING_DISABLING_FACTOR>::disable_limit(
active_set_size_after_slash,
),
);
let session_validators_before = Session::validators();
roll_to_epm_off();
assert!(ElectionProviderMultiPhase::current_phase().is_off());
// Find the indices of the disabled validators
let active_set = Session::validators();
let expected_disabled = slashed
.into_iter()
.map(|d| active_set.iter().position(|a| *a == d).unwrap() as u32)
.collect::<Vec<_>>();
assert_eq!(pallet_staking::ForceEra::<Runtime>::get(), pallet_staking::Forcing::NotForcing);
// slashes so that staking goes into `Forcing::ForceNew`.
slash_through_offending_threshold();
assert_eq!(pallet_staking::ForceEra::<Runtime>::get(), pallet_staking::Forcing::ForceNew);
advance_session_delayed_solution(pool_state.clone());
assert!(ElectionProviderMultiPhase::current_phase().is_emergency());
log_current_time();
let era_before_delayed_next = Staking::current_era();
// try to advance 2 eras.
assert!(start_next_active_era_delayed_solution(pool_state.clone()).is_ok());
assert_eq!(Staking::current_era(), era_before_delayed_next);
assert!(start_next_active_era(pool_state).is_err());
assert_eq!(Staking::current_era(), era_before_delayed_next);
// EPM is still in emergency phase.
assert!(ElectionProviderMultiPhase::current_phase().is_emergency());
// session validator set remains the same.
assert_eq!(Session::validators(), session_validators_before);
// performs recovery through the set emergency result.
let supports = to_supports(&vec![
StakedAssignment { who: 21, distribution: vec![(21, 10)] },
StakedAssignment { who: 31, distribution: vec![(21, 10), (31, 10)] },
StakedAssignment { who: 41, distribution: vec![(41, 10)] },
]);
assert!(ElectionProviderMultiPhase::set_emergency_election_result(
RuntimeOrigin::root(),
supports
)
.is_ok());
// EPM can now roll to signed phase to proceed with elections. The validator set is the
// expected (ie. set through `set_emergency_election_result`).
roll_to_epm_signed();
//assert!(ElectionProviderMultiPhase::current_phase().is_signed());
assert_eq!(Session::validators(), vec![21, 31, 41]);
assert_eq!(Staking::current_era(), era_before_delayed_next.map(|e| e + 1));
assert_eq!(Session::disabled_validators(), expected_disabled);
});
}
@@ -253,77 +225,7 @@ fn continuous_slashes_below_offending_threshold() {
}
#[test]
/// Slashed validator sets intentions in the same era of slashing.
///
/// When validators are slashed, they are chilled and removed from the current `VoterList`. Thus,
/// the slashed validator should not be considered in the next validator set. However, if the
/// slashed validator sets its intention to validate again in the same era when it was slashed and
/// chilled, the validator may not be removed from the active validator set across eras, provided
/// it would selected in the subsequent era if there was no slash. Nominators of the slashed
/// validator will also be slashed and chilled, as expected, but the nomination intentions will
/// remain after the validator re-set the intention to be validating again.
///
/// This behaviour is due to removing implicit chill upon slash
/// <https://github.com/paritytech/substrate/pull/12420>.
///
/// Related to <https://github.com/paritytech/substrate/issues/13714>.
fn set_validation_intention_after_chilled() {
use frame_election_provider_support::SortedListProvider;
use pallet_staking::{Event, Forcing, Nominators};
let (ext, pool_state, _) = ExtBuilder::default()
.epm(EpmExtBuilder::default())
.staking(StakingExtBuilder::default())
.build_offchainify();
execute_with(ext, || {
assert_eq!(active_era(), 0);
// validator is part of the validator set.
assert!(Session::validators().contains(&41));
assert!(<Runtime as pallet_staking::Config>::VoterList::contains(&41));
// nominate validator 81.
assert_ok!(Staking::nominate(RuntimeOrigin::signed(21), vec![41]));
assert_eq!(Nominators::<Runtime>::get(21).unwrap().targets, vec![41]);
// validator is slashed. it is removed from the `VoterList` through chilling but in the
// current era, the validator is still part of the active validator set.
add_slash(&41);
assert!(Session::validators().contains(&41));
assert!(!<Runtime as pallet_staking::Config>::VoterList::contains(&41));
assert_eq!(
staking_events(),
[
Event::Chilled { stash: 41 },
Event::ForceEra { mode: Forcing::ForceNew },
Event::SlashReported {
validator: 41,
slash_era: 0,
fraction: Perbill::from_percent(10)
}
],
);
// after the nominator is slashed and chilled, the nominations remain.
assert_eq!(Nominators::<Runtime>::get(21).unwrap().targets, vec![41]);
// validator sets intention to stake again in the same era it was chilled.
assert_ok!(Staking::validate(RuntimeOrigin::signed(41), Default::default()));
// progress era and check that the slashed validator is still part of the validator
// set.
assert!(start_next_active_era(pool_state).is_ok());
assert_eq!(active_era(), 1);
assert!(Session::validators().contains(&41));
assert!(<Runtime as pallet_staking::Config>::VoterList::contains(&41));
// nominations are still active as before the slash.
assert_eq!(Nominators::<Runtime>::get(21).unwrap().targets, vec![41]);
})
}
#[test]
/// Active ledger balance may fall below ED if account chills before unbonding.
/// Active ledger balance may fall below ED if account chills before unbounding.
///
/// Unbonding call fails if the remaining ledger's stash balance falls below the existential
/// deposit. However, if the stash is chilled before unbonding, the ledger's active balance may
@@ -35,7 +35,7 @@ use sp_runtime::{
transaction_validity, BuildStorage, PerU16, Perbill, Percent,
};
use sp_staking::{
offence::{DisableStrategy, OffenceDetails, OnOffenceHandler},
offence::{OffenceDetails, OnOffenceHandler},
EraIndex, SessionIndex,
};
use sp_std::prelude::*;
@@ -236,7 +236,6 @@ parameter_types! {
pub const SessionsPerEra: sp_staking::SessionIndex = 2;
pub static BondingDuration: sp_staking::EraIndex = 28;
pub const SlashDeferDuration: sp_staking::EraIndex = 7; // 1/4 the bonding duration.
pub const OffendingValidatorsThreshold: Perbill = Perbill::from_percent(40);
pub HistoryDepth: u32 = 84;
}
@@ -290,6 +289,8 @@ parameter_types! {
/// Upper limit on the number of NPOS nominations.
const MAX_QUOTA_NOMINATIONS: u32 = 16;
/// Disabling factor set explicitly to byzantine threshold
pub(crate) const SLASHING_DISABLING_FACTOR: usize = 3;
impl pallet_staking::Config for Runtime {
type Currency = Balances;
@@ -308,7 +309,6 @@ impl pallet_staking::Config for Runtime {
type EraPayout = ();
type NextNewSession = Session;
type MaxExposurePageSize = ConstU32<256>;
type OffendingValidatorsThreshold = OffendingValidatorsThreshold;
type ElectionProvider = ElectionProviderMultiPhase;
type GenesisElectionProvider = onchain::OnChainExecution<OnChainSeqPhragmen>;
type VoterList = BagsList;
@@ -320,6 +320,7 @@ impl pallet_staking::Config for Runtime {
type EventListeners = Pools;
type WeightInfo = pallet_staking::weights::SubstrateWeight<Runtime>;
type BenchmarkingConfig = pallet_staking::TestBenchmarkingConfig;
type DisablingStrategy = pallet_staking::UpToLimitDisablingStrategy<SLASHING_DISABLING_FACTOR>;
}
impl<LocalCall> frame_system::offchain::SendTransactionTypes<LocalCall> for Runtime
@@ -871,7 +872,6 @@ pub(crate) fn on_offence_now(
offenders,
slash_fraction,
Staking::eras_start_session_index(now).unwrap(),
DisableStrategy::WhenSlashed,
);
}
@@ -886,19 +886,16 @@ pub(crate) fn add_slash(who: &AccountId) {
);
}
// Slashes enough validators to cross the `Staking::OffendingValidatorsThreshold`.
pub(crate) fn slash_through_offending_threshold() {
let validators = Session::validators();
let mut remaining_slashes =
<Runtime as pallet_staking::Config>::OffendingValidatorsThreshold::get() *
validators.len() as u32;
// Slashes 1/2 of the active set. Returns the `AccountId`s of the slashed validators.
pub(crate) fn slash_half_the_active_set() -> Vec<AccountId> {
let mut slashed = Session::validators();
slashed.truncate(slashed.len() / 2);
for v in validators.into_iter() {
if remaining_slashes != 0 {
add_slash(&v);
remaining_slashes -= 1;
}
for v in slashed.iter() {
add_slash(v);
}
slashed
}
// Slashes a percentage of the active nominators that haven't been slashed yet, with