feat(sequencer): set block building limits from checkpoint limits by spalladino · Pull Request #20974 · AztecProtocol/aztec-packages

spalladino · 2026-02-27T21:04:15Z

The checkpoint builder now tracks remaining L2 gas, DA gas, and blob fields in a checkpoint while building each block, and forwards them to the public processor. This means that a proposer will not propose blocks that overall exceed checkpoint limits, and validators will properly reject them.

In addition, the proposer defaults the L2 and DA gas limits per block to the checkpoint limits divided by expected number of blocks, times two. This value is still capped by the remaining gas in the checkpoint builder, but means that a proposer will not waste the entire checkpoint gas allocation on the first block.

Fixes A-528

Changes

As described by Claude

Derives per-block L2 and DA gas budgets from L1 checkpoint limits (rollupManaLimit, MAX_PROCESSABLE_DA_GAS_PER_CHECKPOINT) using the timetable's maxNumberOfBlocks and a configurable multiplier (default: 2) for creating proposals only (not for validation)
Moves blob field and gas tracking from the checkpoint proposal job loop into CheckpointBuilder.capLimitsByCheckpointBudgets(), which caps per-block limits by remaining checkpoint budgets for both proposer and validator paths
Plumbs maxTxsPerBlock, maxL2BlockGas, maxDABlockGas, and rollupManaLimit through to validator re-execution so limits are enforced during block validation
Replaces byte-based maxBlockSizeInBytes with field-based blob limits and a pre-processing blob field estimation (getPrivateTxEffectsSizeInFields)
Adds gasPerBlockAllocationMultiplier config (SEQ_GAS_PER_BLOCK_ALLOCATION_MULTIPLIER, default: 2)
Makes maxL2BlockGas and maxDABlockGas optional (auto-computed from checkpoint limits when not set)

🤖 Generated with Claude Code

spalladino · 2026-02-27T21:17:54Z

yarn-project/validator-client/src/checkpoint_builder.ts

+    // Remaining L2 gas (mana)
+    // IMPORTANT: This assumes mana is computed solely based on L2 gas used in transactions.
+    // This may change in the future.
+    const usedMana = sum(existingBlocks.map(b => b.header.totalManaUsed.toNumber()));
+    const remainingMana = this.config.rollupManaLimit - usedMana;
+
+    // Remaining DA gas: DA gas = tx blob fields * DA_GAS_PER_FIELD
+    // IMPORTANT: This assumes DA gas is computed solely based on the number of blob fields in transactions
+    // This may change in the future, but we cannot access the actual DA gas used in a block since it's not exposed
+    // in the L2BlockHeader, so we have to rely on recomputing it here.
+    const usedDAGas =
+      sum(existingBlocks.map(b => sum(b.body.txEffects.map(tx => tx.getNumBlobFields())))) * DA_GAS_PER_FIELD;
+    const remainingDAGas = MAX_PROCESSABLE_DA_GAS_PER_CHECKPOINT - usedDAGas;
+
+    // Remaining blob fields (block blob fields include both tx data and block-end overhead)
+    const usedBlobFields = sum(existingBlocks.map(b => b.toBlobFields().length));
+    const totalBlobCapacity = BLOBS_PER_CHECKPOINT * FIELDS_PER_BLOB - NUM_CHECKPOINT_END_MARKER_FIELDS;
+    const isFirstBlock = existingBlocks.length === 0;
+    const blockEndOverhead = getNumBlockEndBlobFields(isFirstBlock);
+    const maxBlobFieldsForTxs = totalBlobCapacity - usedBlobFields - blockEndOverhead;
+
+    // Cap L2 gas by remaining checkpoint mana
+    const cappedL2Gas = Math.min(opts.maxBlockGas?.l2Gas ?? remainingMana, remainingMana);
+
+    // Cap DA gas by remaining checkpoint DA gas budget
+    const cappedDAGas = Math.min(opts.maxBlockGas?.daGas ?? remainingDAGas, remainingDAGas);
+
+    // Cap blob fields by remaining checkpoint blob capacity
+    const cappedBlobFields =
+      opts.maxBlobFields !== undefined ? Math.min(opts.maxBlobFields, maxBlobFieldsForTxs) : maxBlobFieldsForTxs;
+
+    return {
+      maxBlockGas: new Gas(cappedDAGas, cappedL2Gas),
+      maxBlobFields: cappedBlobFields,
+    };


@sirasistant can I ask you to review this part of the code?

It looks good, however the cappedDAGas is a bit pointless since we'd always run out of maxBlobFields first (since DA gas doesn't account block builder blob fields overheads)

since DA gas doesn't account block builder blob fields overheads

Should it? I'd feel more comfortable if we could remove the max-blob-fields check eventually and just rely on DA gas.

Problem is that some fields don't come from txs: come from builder decided things like how many blocks you create in one checkpoint. We'd need blocks to increase da gas used and have da gas used in the header. Also it'd remove the ability for us (which we don't exercise now) of overcharging DA gas for long term storage for example (making the DA gas of nullifiers higher, for example) So i'm not sure it should

spalladino · 2026-02-27T22:09:44Z

yarn-project/simulator/src/public/public_processor/public_processor.ts

@sirasistant also the changes on this file

spalladino · 2026-02-27T22:38:48Z

noir-projects/noir-protocol-circuits/crates/types/src/constants.nr


 // Default gas limits. Users should use gas estimation, or they will overpay gas fees.
-// TODO: consider moving to typescript
+// TODO: These are overridden in typescript-land. Remove them from here.


I didn't delete them in this PR because I was scared of triggering a change in a vk

If they are not used in any circuit / protocol contract, it shouldn't trigger a vk change

… building During re-execution (validation/proving), the public processor must process the exact txs from the proposal. Pre-processing skip checks for estimated blob fields and gas limits are now gated behind a new `isBuildingProposal` flag on PublicProcessorLimits, which is only set by the sequencer. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

BEGIN_COMMIT_OVERRIDE fix: track last seen nonce in case of stale fallback L1 RPC node (#20855) feat: Validate num txs in block proposals (#20850) fix(archiver): enforce checkpoint boundary on rollbackTo (#20908) fix: tps zero metrics (#20656) fix: handle scientific notation in bigintConfigHelper (#20929) feat(aztec): node enters standby mode on genesis root mismatch (#20938) fix: logging of class instances (#20807) feat(slasher): make slash grace period relative to rollup upgrade time (#20942) chore: add script to find PRs to backport (#20956) chore: remove unused prover-node dep (#20955) fix: increase minFeePadding in e2e_bot bridge resume tests and harden GasFees.mul() (#20962) feat(sequencer): (A-526) rotate publishers when send fails (#20888) chore: (A-554) bump reth version 1.6.0 -> 1.11.1 for eth devnet (#20889) chore: metric on how many epochs validator has been on committee (#20967) fix: set wallet minFeePadding in BotFactory constructor (#20992) chore: deflake epoch invalidate block test (#21001) chore(sequencer): e2e tests for invalid signature recovery in checkpoint attestations (#20971) chore: deflake duplicate proposals and attestations (#20990) chore: deflake epochs mbps test (#21003) feat: reenable function selectors in txPublicSetupAllowList (#20909) fix: limit offenses when voting in tally slashing mode by slashMaxPayloadSize (#20683) fix(spartan): wire SEQ_L1_PUBLISHING_TIME_ALLOWANCE_IN_SLOT env var (#21017) END_COMMIT_OVERRIDE

PhilWindle · 2026-03-03T10:37:36Z

yarn-project/sequencer-client/src/client/sequencer-client.ts

    const l1PublishingTimeBasedOnChain = isAnvilTestChain(config.l1ChainId) ? 1 : ethereumSlotDuration;
    const l1PublishingTime = config.l1PublishingTime ?? l1PublishingTimeBasedOnChain;

+    const { maxL2BlockGas, maxDABlockGas } = this.computeBlockGasLimits(config, rollupManaLimit, l1PublishingTime, log);


So we take the combination of rollup defined limits and user provided config to produce specific configuration to be passed to the sequencer.

Not just that. We take user-defined block-level limits, along with checkpoint-level limits (which are sourced from either the Rollup contract, constants, or more user config), and derive the final block-level limits to be used, and feed them into the sequencer.

PhilWindle · 2026-03-03T10:38:52Z

yarn-project/sequencer-client/src/sequencer/checkpoint_proposal_job.ts

-      const maxBlobFieldsForTxs = remainingBlobFields - blockEndOverhead;
-
+      // Note that gas and blob field limits are further capped by checkpoint-level budgets inside CheckpointBuilder
      const blockBuilderOptions: PublicProcessorLimits = {


And that configuration is what we are looking at here.

PhilWindle · 2026-03-03T10:43:15Z

yarn-project/sequencer-client/src/config.ts

-    ...numberConfigHelper(DefaultSequencerConfig.maxDABlockGas),
+    parseEnv: (val: string) => (val ? parseInt(val, 10) : undefined),
+  },
+  gasPerBlockAllocationMultiplier: {


Is this to allow sequencers to overfill earlier blocks if they want to, but then they would need to ease off in later blocks to adhere to the checkpoint limit defined on the rollup?

Correct. It's also there because different blocks may hit different limits (l2 gas, da gas, blob fields, tx count). So if we limited each block by exactly checkpoint_limit / num_blocks on all dimensions, we'd be wasting a lot of checkpoint capacity.

PhilWindle · 2026-03-03T10:55:46Z

yarn-project/stdlib/src/tx/tx.ts

-    );
+  getPrivateTxEffectsSizeInFields(): number {
+    // 3 fields overhead: tx_start_marker, tx_hash, tx_fee
+    const overheadFields = TX_DA_GAS_OVERHEAD / DA_GAS_PER_FIELD;


Does this need a Math.ceil?

Not really, since TX_DA_GAS_OVERHEAD is defined as:

pub global TX_DA_GAS_OVERHEAD: u32 = 3 * DA_GAS_PER_FIELD

aztec-packages/noir-projects/noir-protocol-circuits/crates/types/src/constants.nr

Line 1039 in f3b702f

pub global TX_DA_GAS_OVERHEAD: u32 = 3 * DA_GAS_PER_FIELD;

PhilWindle · 2026-03-03T10:57:57Z

yarn-project/validator-client/src/block_proposal_handler.ts

+      // set if the operator explicitly configures maxL2BlockGas/maxDABlockGas.
+      // Checkpoint-level caps are still enforced inside CheckpointBuilder.buildBlock via
+      // capLimitsByCheckpointBudgets, so L1 rejection is prevented even without per-block limits.
+      maxBlockGas:


Not sure I follow why we are setting any limits here?

I was on the fence on whether to add it or not.

Rationale was that if we want to set a hard limit to (let's say) L2 gas per block (independent of the accumulated value across the checkpoint), then if a sequencer sent a block with more than that, it wouldn't be considered valid. But it's an easy way to brick things.

I'll remove it for now, and we can discuss later how we want to handle it. Options are:

Validators only check checkpoint-level limits, not block-level ones. Simplest option, but it means the network cannot enforce any block-level limit.

Validators check block-level limits as well. This means that if a single operator wants to produce slightly larger blocks than what's configured per-block, they won't get attested to. This also means that gradually increasing block size (while staying within checkpoint limits) is more difficult.

We have separate config for block-level limits for proposal and validation. Most complex option, but allows the greatest deal of flexibility.

Created A-613 to track

…ild-limits

See A-613

AztecBot · 2026-03-03T13:53:53Z

❌ Failed to cherry-pick to v4 due to conflicts. Dispatching ClaudeBox to resolve. View backport run.

…ckport #20974) (#21055) ## Summary Backport of #20974 to v4. The checkpoint builder now tracks remaining L2 gas, DA gas, and blob fields in a checkpoint while building each block, and forwards them to the public processor. This means that a proposer will not propose blocks that overall exceed checkpoint limits, and validators will properly reject them. **Cherry-picked commits:** - feat(sequencer): set block building limits from checkpoint limits - feat(sequencer): only skip txs due to gas/blob limits during proposal building - fix(gas): saner defaults for da gas limit per tx - fix: fix spread when computing processor limits - fix(processor): check gas limits using used gas on reexecution - test: check that validators reject exceeded gas limits - chore: re-validate checkpoint limits before proposal and attestation - test: fix tests related to avm gas limits - fix(validator): do not check block-level limits on validation - chore: add inline comments **Conflict resolution:** - `timetable.ts`: Resolved log level conflict — v4 uses `createLogger` (always defined) so kept `this.log.info(` without optional chaining (matching PR's intent to upgrade from `verbose` to `info`). ClaudeBox log: http://ci.aztec-labs.com/2433a99083562757-1 --------- Co-authored-by: Santiago Palladino <santiago@aztec-labs.com> Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com> Co-authored-by: Santiago Palladino <santiago@aztecprotocol.com>

BEGIN_COMMIT_OVERRIDE test: update proving-real test to mbps (#20991) chore: epoch proving log analyzer (#21033) chore: update pause script to allow resume (#21032) feat: price bump for RPC transaction replacement (#20806) refactor: remove update checker, retain version checks (#20898) fix: (A-592) p2p client proposal tx collector test (#20998) refactor: use publishers-per-pod in deployments (#21039) chore: web3signer refreshes keystore (#21045) feat(sequencer): set block building limits from checkpoint limits (#20974) chore(e2e): fix e2e bot L1 tx nonce reuse (#21052) feat: Update L1 to L2 message APIs (#20913) fix: (A-589) epochs l1 reorgs test (#20999) feat(sequencer): add SEQ_MAX_TX_PER_CHECKPOINT config (#21016) fix: drop --pid=host from docker_isolate (#21081) feat: standby mode for prover broker (#21098) fix(p2p): remove default block handler in favor of block handler (#21105) feat(validator): add VALIDATOR_ env vars for independent block limits (#21060) refactor(p2p): decouple proposal validators from base class via composition (#21075) feat: additional validation in public setup allowlist (onlySelf + null msg sender) (#21122) fix: (A-591) aztecProofSubmissionEpochs incorrectly named as aztecProofSubmissionWindow (#21108) refactor(sequencer): rename SEQ_GAS_PER_BLOCK_ALLOCATION_MULTIPLIER to SEQ_PER_BLOCK_ALLOCATION_MULTIPLIER (#21125) fix: unbound variable in check_doc_references.sh with set -u (#21126) feat: calldata length validation of public setup function allowlist (#21139) fix: include mismatched values in tx metadata validation errors (#21147) feat: single-node implementation of slash-protection signer (#20894) feat: Remove non-protocol contracts from public setup allowlist (#21154) chore: More updated Alpha configuration (#21155) chore: tally slashing pruning improvements (#21161) fix: update dependencies (#20997) fix: omit bigint priceBumpPercentage from IPC config in testbench worker (#21169) refactor(p2p): (A-588) maintain sorted array in tx pool instead of sorting on read (#21079) fix(p2p): report most severe failure in runValidations (#21185) fix: use dedicated L1 account for bot bridge resume tests to avoid nonce race (#21148) fix: parse error.message in formatViemError (#21163) fix: bump lighthouse consensus client v7.1.0 -> v8.0.1 (#21170) chore: code decuplication + refactor (public setup allowlist) (#21200) END_COMMIT_OVERRIDE

spalladino requested review from dbanks12 and fcarreiro as code owners February 27, 2026 21:04

spalladino added ci-no-fail-fast Sets NO_FAIL_FAST in the CI so the run is not aborted on the first failure backport-to-v4 labels Feb 27, 2026

spalladino requested a review from sirasistant as a code owner February 27, 2026 21:04

spalladino added ci-no-fail-fast Sets NO_FAIL_FAST in the CI so the run is not aborted on the first failure backport-to-v4 labels Feb 27, 2026

spalladino removed request for dbanks12, fcarreiro and sirasistant February 27, 2026 21:09

spalladino commented Feb 27, 2026

View reviewed changes

spalladino requested a review from LeilaWang as a code owner February 27, 2026 21:59

spalladino removed the request for review from LeilaWang February 27, 2026 22:09

spalladino commented Feb 27, 2026

View reviewed changes

spalladino and others added 7 commits March 2, 2026 09:34

feat(sequencer): set block building limits from checkpoint limits

cfe074b

fix(gas): saner defaults for da gas limit per tx

65779d5

fix: fix spread when computing processor limits

a88619c

fix(processor): check gas limits using used gas on reexecution

21b54b8

test: check that validators reject exceeded gas limits

51235f7

chore: re-validate checkpoint limits before proposal and attestation

e2e6d00

spalladino force-pushed the palla/update-block-build-limits branch from 0cdd6ac to 0e27a6c Compare March 2, 2026 16:23

spalladino requested review from IlyasRidhuan and jeanmon as code owners March 2, 2026 16:23

spalladino force-pushed the palla/update-block-build-limits branch from 0e27a6c to b0c9e08 Compare March 2, 2026 16:25

test: fix tests related to avm gas limits

f3b702f

spalladino force-pushed the palla/update-block-build-limits branch from b0c9e08 to f3b702f Compare March 2, 2026 16:51

PhilWindle reviewed Mar 3, 2026

View reviewed changes

spalladino added 3 commits March 3, 2026 09:48

Merge remote-tracking branch 'origin/next' into palla/update-block-bu…

88704e6

…ild-limits

fix(validator): do not check block-level limits on validation

a6c1d7a

See A-613

chore: add inline comments

1dedb44

PhilWindle approved these changes Mar 3, 2026

View reviewed changes

spalladino merged commit bfa7fa8 into merge-train/spartan Mar 3, 2026
13 of 15 checks passed

spalladino deleted the palla/update-block-build-limits branch March 3, 2026 13:49

AztecBot mentioned this pull request Mar 3, 2026

feat: merge-train/spartan #21034

Merged

AztecBot mentioned this pull request Mar 3, 2026

feat(sequencer): set block building limits from checkpoint limits (backport #20974) #21055

Merged

AztecBot mentioned this pull request Mar 3, 2026

chore(v4): release 4.1.1 #20979

Open

Conversation

spalladino commented Feb 27, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Changes

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

spalladino Feb 27, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

spalladino Feb 27, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

spalladino Mar 3, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

AztecBot commented Mar 3, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

spalladino commented Feb 27, 2026 •

edited

Loading

spalladino Feb 27, 2026 •

edited

Loading

spalladino Feb 27, 2026 •

edited

Loading

spalladino Mar 3, 2026 •

edited

Loading