Update Earliest available slot when pruning #15694

satushh · 2025-09-15T12:04:03Z

What type of PR is this?

Bug fix

What does this PR do? Why is it needed?

earliest_available_slot needs to be updated (if required) when pruning occurs. Since peers will think we have data for slot n, while in reality, we wiped data up to slot m with n < m.

Which issues(s) does this PR fix?

Part of #14129

Other notes for review

Steps to reproduce:

Clone the branches from these PRs:
- add: support for MIN_EPOCHS_FOR_BLOCK_REQUESTS customization ethpandaops/ethereum-genesis-generator#242
- feat: add support for MIN_EPOCHS_FOR_BLOCK_REQUESTS ethpandaops/ethereum-package#1211
In the ethereum-genesis-generator folder, run: docker build -t ethereum-genesis-generator:local .
Build local prysm:

bazel build //cmd/beacon-chain:oci_image_tarball \
  --platforms=@io_bazel_rules_go//go/toolchain:linux_arm64_cgo \
  --config=release \
  --//proto:network=minimal \
  --@io_bazel_rules_go//go/config:tags=minimal

docker load -i bazel-bin/cmd/beacon-chain/oci_image_tarball/tarball.tar
docker tag gcr.io/offchainlabs/prysm/beacon-chain prysm-bn-custom-image:minimal-build

bazel build //cmd/validator:oci_image_tarball \
  --platforms=@io_bazel_rules_go//go/toolchain:linux_arm64_cgo \
  --config=release \
  --//proto:network=minimal \
  --@io_bazel_rules_go//go/config:tags=minimal

docker load -i bazel-bin/cmd/validator/oci_image_tarball/tarball.tar
docker tag gcr.io/offchainlabs/prysm/validator prysm-vc-custom-image:minimal-build

Then use this yaml in ethereum-package to run kurtosis:

participants:
  - el_type: geth
    el_image: ethpandaops/geth:fusaka-devnet-3
    cl_type: prysm
    cl_image: "prysm-bn-custom-image:minimal-build"
    vc_image: "prysm-vc-custom-image:minimal-build"
    count: 2
    supernode: true
    cl_extra_params:
      - --verbosity=debug
      - --pruner-retention-epochs=4
      - --beacon-db-pruning
      - --subscribe-all-subnets
network_params:
  preset: minimal
  electra_fork_epoch: 0
  fulu_fork_epoch: 1
  min_epochs_for_block_requests: 3
ethereum_genesis_generator_params:
  image: ethereum-genesis-generator:local
additional_services:
  - dora        
  - spamoor
spamoor_params:
  image: ethpandaops/spamoor:master
  max_mem: 4000
  spammers:
    - scenario: eoatx
      config:
        throughput: 200
    - scenario: blobs
      config:
        throughput: 20

Acknowledgements

I have read CONTRIBUTING.md.
I have included a uniquely named changelog fragment file.
I have added a description to this PR with sufficient context for reviewers to understand this PR.

Copilot

Pull Request Overview

Updates the beacon chain node to properly maintain the earliest available slot value when pruning operations occur. This ensures peers receive accurate information about data availability after pruning removes older blocks and data columns.

Integrates custody updater functionality with both pruner and data column storage services
Modifies pruning logic to update earliest available slot after data removal
Adds comprehensive test coverage for custody update scenarios

Reviewed Changes

Copilot reviewed 8 out of 8 changed files in this pull request and generated 3 comments.

Show a summary per file

File	Description
changelog/satushh-update-easlot-pruning.md	Documents the feature in changelog
beacon-chain/node/node.go	Configures P2P service as custody updater for pruner and data column storage
beacon-chain/db/pruner/pruner.go	Adds custody updater interface and updates earliest available slot after pruning
beacon-chain/db/pruner/pruner_test.go	Tests custody update functionality with mock implementations
beacon-chain/db/filesystem/data_column.go	Adds custody updater support to data column storage pruning
beacon-chain/db/filesystem/data_column_test.go	Tests data column storage custody updates
beacon-chain/db/pruner/BUILD.bazel	Adds test dependency for assert library
beacon-chain/db/filesystem/BUILD.bazel	Adds test dependency for assert library

_{Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.}

Copilot · 2025-09-16T11:14:43Z

beacon-chain/db/pruner/pruner.go

+		custodyGroupCount, err := p.custody.CustodyGroupCount()
+		if err != nil {
+			log.WithError(err).Error("Failed to get custody group count, cannot update earliest available slot after pruning")
+			return nil


Returning nil here prevents the method from properly signaling an error state to callers. This should return the original error instead of nil since the custody update failure represents a genuine error condition that calling code should be aware of.

Suggested change

return nil

return errors.Wrap(err, "failed to get custody group count for earliest available slot update")

But the pruning itself is successful.

Copilot · 2025-09-16T11:14:44Z

beacon-chain/db/filesystem/data_column.go

+			return
+		}
+
+		// Get current custody group count to preserve it during update
+		custodyGroupCount, err := dcs.custody.CustodyGroupCount()
+		if err != nil {
+			log.WithError(err).Error("Failed to get custody group count, cannot update earliest available slot after data column pruning")
+			return


Early returns from this method prevent the cleanup of muChans map which happens after the custody update logic. The custody update errors should be logged but not cause early returns, or the muChans cleanup should be moved before the custody update logic.

Suggested change

return

}

// Get current custody group count to preserve it during update

custodyGroupCount, err := dcs.custody.CustodyGroupCount()

if err != nil {

log.WithError(err).Error("Failed to get custody group count, cannot update earliest available slot after data column pruning")

return

// Do not return early; continue to allow all cleanup logic to run

}

// Get current custody group count to preserve it during update

custodyGroupCount, err := dcs.custody.CustodyGroupCount()

if err != nil {

log.WithError(err).Error("Failed to get custody group count, cannot update earliest available slot after data column pruning")

// Do not return early; continue to allow all cleanup logic to run

Not sure what custodyGroupCount should be set to here if we can't fetch it.

Copilot · 2025-09-16T11:14:44Z

beacon-chain/db/filesystem/data_column.go

+			return
+		}
+
+		// Get current custody group count to preserve it during update
+		custodyGroupCount, err := dcs.custody.CustodyGroupCount()
+		if err != nil {
+			log.WithError(err).Error("Failed to get custody group count, cannot update earliest available slot after data column pruning")
+			return


Early returns from this method prevent the cleanup of muChans map which happens after the custody update logic. The custody update errors should be logged but not cause early returns, or the muChans cleanup should be moved before the custody update logic.

Suggested change

return

}

// Get current custody group count to preserve it during update

custodyGroupCount, err := dcs.custody.CustodyGroupCount()

if err != nil {

log.WithError(err).Error("Failed to get custody group count, cannot update earliest available slot after data column pruning")

return

// Do not return early; continue to cleanup

}

// Get current custody group count to preserve it during update

custodyGroupCount, err := dcs.custody.CustodyGroupCount()

if err != nil {

log.WithError(err).Error("Failed to get custody group count, cannot update earliest available slot after data column pruning")

// Do not return early; continue to cleanup

prestonvanloon · 2025-09-16T16:13:52Z

changelog/satushh-update-easlot-pruning.md

@@ -0,0 +1,3 @@
+### Added
+
+- Update the earliest available slot while pruning. 


Could you add a bit more to this changelog? These are used to compile the release notes and this isn't enough context to understand why this change is important

prestonvanloon · 2025-09-16T16:16:37Z

beacon-chain/db/pruner/pruner.go

+	if p.custody != nil {
+		earliestAvailableSlot := pruneUpto + 1
+
+		// Get current custody group count to preserve it during update
+		custodyGroupCount, err := p.custody.CustodyGroupCount()
+		if err != nil {
+			log.WithError(err).Error("Failed to get custody group count, cannot update earliest available slot after pruning")
+			return nil
+		}
+
+		// Update the custody info with new earliest available slot
+		_, _, err = p.custody.UpdateCustodyInfo(earliestAvailableSlot, custodyGroupCount)
+		if err != nil {
+			log.WithError(err).WithField("earliestAvailableSlot", earliestAvailableSlot).
+				Error("Failed to update earliest available slot after pruning")
+		} else {
+			log.WithField("earliestAvailableSlot", earliestAvailableSlot).
+				Debug("Updated earliest available slot after pruning")
+		}
+	}


All of this should probably into its own method for a better abstraction/separation of concerns.

Suggested change

if p.custody != nil {

earliestAvailableSlot := pruneUpto + 1

// Get current custody group count to preserve it during update

custodyGroupCount, err := p.custody.CustodyGroupCount()

if err != nil {

log.WithError(err).Error("Failed to get custody group count, cannot update earliest available slot after pruning")

return nil

}

// Update the custody info with new earliest available slot

_, _, err = p.custody.UpdateCustodyInfo(earliestAvailableSlot, custodyGroupCount)

if err != nil {

log.WithError(err).WithField("earliestAvailableSlot", earliestAvailableSlot).

Error("Failed to update earliest available slot after pruning")

} else {

log.WithField("earliestAvailableSlot", earliestAvailableSlot).

Debug("Updated earliest available slot after pruning")

}

}

p.updateEarliestSlot(pruneUpto+1)

prestonvanloon · 2025-09-16T16:19:03Z

beacon-chain/db/filesystem/data_column.go

+
+	// Update the earliest available slot via injected updater after pruning.
+	// The earliest available slot is the first slot after the pruned epochs.
+	if dcs.custody != nil {


Same feedback here about moving the "update earliest slot" logic out of the prune method and into its own method.

…Fulu

nalepae · 2025-09-25T14:44:20Z

beacon-chain/db/pruner/pruner_test.go

+			require.NoError(t, err)
+
+			// Test the pruning calculation
+			currentSlot := primitives.Slot(20000) // Reasonable slot number for minimal config


Don't assume this kind of "reasonable" number. If the ETH spec changes, then this test may break.
Instead, override the config.MinEpochsForBlockRequests at the start of the test.

Unfortunatly, for some historical reasons, this value is recomputed in the MinEpochsForBlockRequests function.

This function was defined and used before the corresponding value was introduced in the spec.
So, IMO the cleanest way to do it is:

In the Prysm codebase, remove the MinEpochsForBlockRequests and directly use config.MinEpochsForBlockRequests where MinEpochsForBlockRequests where used.

In this test, override config.MinEpochsForBlockRequests

Addressed here: f324089

nalepae · 2025-10-13T09:54:06Z

beacon-chain/node/node.go

 	}

+	// P2P service is required for pruner to update earliest available slot after pruning
+	p2pService := b.fetchP2P()


This check can be done directly in pruner.New, so responsability is shifter from caller to callee.

nalepae · 2025-10-13T11:42:45Z

beacon-chain/db/kv/custody.go


-// UpdateCustodyInfo atomically updates the custody group count only it is greater than the stored one.
-// In this case, it also updates the earliest available slot with the provided value.
+// UpdateCustodyInfo updates the custody group count if it is greater than the stored one,


Why atomically had been removed?

nalepae · 2025-10-13T11:45:13Z

beacon-chain/db/kv/custody.go

+		// Update earliest available slot if it advanced.
+		// This is for pruning to work correctly as blocks are pruned,
+		// the earliest available slot moves forward independently of custody group changes.
+		if earliestAvailableSlot > storedEarliestAvailableSlot {


Increasing the earliestAvailableSlot only while not increasing the custodyGroupCount at the same time should not be allowed is the new value of earliestAvailableSlot is higher than the minimum slot to be stored.

Else it means that a node could simply choose to put earliestAvailableslot = currentSlot and do not serve any data to its peers.

Same comment as #15694 (comment) actually.

nalepae · 2025-10-13T11:48:18Z

beacon-chain/db/pruner/pruner.go

+// custodyUpdater is a tiny interface that p2p service implements; kept here to avoid
+// importing the p2p package and creating a cycle.
+type custodyUpdater interface {
+	CustodyGroupCount(context.Context) (uint64, error)


CustodyGroupCount can be renamed into GroupCount to avoid

...custody.CutodyGroupCount

nalepae · 2025-10-13T11:49:57Z

beacon-chain/db/pruner/pruner.go

+
+// updateEarliestSlot updates the earliest available slot via the injected custody updater
+// and also persists it to the database.
+func (p *Service) updateEarliestSlot(earliestAvailableSlot primitives.Slot) error {


updateEarliestSlot ==> updateEarliestAvailableSlot to stick with the spec.

This function should have the same checks/protections as UpdateEarliestAvailableSlot on the P2P side.

nalepae · 2025-10-13T11:50:42Z

beacon-chain/db/pruner/pruner.go

+		"duration":              time.Since(tt),
+		"currentSlot":           slot,
+		"batchSize":             defaultPrunableBatchSize,
+		"numBatches":            numBatches,


This log should be displayed after updating the earliest available slot to reflect the real state of the node.

nalepae · 2025-10-13T11:59:08Z

beacon-chain/p2p/custody.go

+	var minRequiredEpoch primitives.Epoch
+	if currentEpoch > minEpochsForBlocks {
+		minRequiredEpoch = currentEpoch - minEpochsForBlocks
+	} else {
+		minRequiredEpoch = 0
+	}


Suggested change

var minRequiredEpoch primitives.Epoch

if currentEpoch > minEpochsForBlocks {

minRequiredEpoch = currentEpoch - minEpochsForBlocks

} else {

minRequiredEpoch = 0

}

minRequiredEpoch := primitives.Epoch(0)

if currentEpoch > minEpochsForBlocks {

minRequiredEpoch = currentEpoch - minEpochsForBlocks

}

nalepae · 2025-10-13T12:02:23Z

beacon-chain/db/pruner/pruner.go

+
+// updateEarliestSlot updates the earliest available slot via the injected custody updater
+// and also persists it to the database.
+func (p *Service) updateEarliestSlot(earliestAvailableSlot primitives.Slot) error {


This function should have the same checks/protections as UpdateEarliestAvailableSlot on the P2P side.

nalepae · 2025-10-13T12:07:10Z

beacon-chain/p2p/custody_test.go

+	params.OverrideBeaconConfig(config)
+
+	t.Run("Valid update", func(t *testing.T) {
+		const initialSlot primitives.Slot = 50


Better to use

const ( initialSlot primitives.Slot = 50 const newSlot primitives.Slot = 100 ... ) grouped declaration.

nalepae · 2025-10-13T12:11:30Z

beacon-chain/db/pruner/pruner.go

+	}
+
+	// Update the earliest available slot
+	if err := p.custody.UpdateEarliestAvailableSlot(earliestAvailableSlot); err != nil {


p.custody.UpdateEarliestAvailableSlot could return the (possibly updated) custody group count to avoid to call p.custody.CustodyGroupCount later, before saving into the DB.

(As done in UpdateCustodyInfo.)

…iestSlot. UpdateEarliestAvailableSlot returns cgc

…lso increases

…vailableSlot returns cgc now

nalepae · 2025-10-16T16:37:19Z

beacon-chain/p2p/custody.go

+	defer s.custodyInfoLock.Unlock()
+
+	if s.custodyInfo == nil {
+		return 0, 0, errors.New("no custody info available")


Nit: Add a test case for that.

nalepae · 2025-10-17T08:36:28Z

beacon-chain/db/kv/custody.go

+		// required slot (based on MIN_EPOCHS_FOR_BLOCK_REQUESTS from current time).
+		// This prevents nodes from arbitrarily refusing to serve mandatory historical data.
+		if earliestAvailableSlot > storedEarliestAvailableSlot {
+			// If custody group count is NOT increasing, validate the increase is allowed


It's better to invert the if condition (and return in the if) to reduce the indentation.

What happens in case of backfill? Is the stored earliestAvailableSlot value decreased in the DB?

…tAvailableSlot) & undo changes made to UpdateCustodyInfo

satushh added 6 commits September 15, 2025 13:02

Update Earliest available slot when pruning

19bb4e6

bazel run //:gazelle -- fix

d355f92

custodyUpdater interface to avoid import cycle

2b055f0

bazel run //:gazelle -- fix

6066724

Merge branch 'develop' into update-easlot-pruning

c69b772

simplify test

c09aba6

satushh requested a review from Copilot September 16, 2025 11:13

Copilot AI reviewed Sep 16, 2025

View reviewed changes

satushh marked this pull request as ready for review September 16, 2025 11:23

satushh requested a review from nalepae September 16, 2025 11:23

prestonvanloon reviewed Sep 16, 2025

View reviewed changes

satushh added 3 commits September 17, 2025 10:21

separation of concerns

3050e44

Merge branch 'develop' into update-easlot-pruning

2e929dd

Merge branch 'develop' into update-easlot-pruning

7ed865a

james-prysm added peer-das fulu optimization PR making Prysm more effective labels Sep 18, 2025

satushh added 12 commits September 19, 2025 17:32

debug log for updating eas

6b04d67

Merge branch 'develop' into update-easlot-pruning

11ebef8

UpdateEarliestAvailableSlot function in CustodyManager

8e24673

Merge branch 'develop' into update-easlot-pruning

7152fa3

fix test

9db57e3

UpdateEarliestAvailableSlot function for FakeP2P

b4364d4

lint

c5b88f1

Merge branch 'develop' into update-easlot-pruning

9f3fdfe

UpdateEarliestAvailableSlot instead of UpdateCustodyInfo + check for …

640aea5

…Fulu

fix test and lint

c731400

Merge branch 'develop' into update-easlot-pruning

fd02744

bugfix: enforce minimum retention period in pruner

55cfe72

nalepae reviewed Sep 25, 2025

View reviewed changes

satushh added 8 commits October 3, 2025 16:59

update the value stored in the DB

45a4edd

log tidy up

9be9cde

use errNoCustodyInfo

f559d11

Merge branch 'develop' into update-easlot-pruning

3d2b002

allow earliestAvailableSlot edit when custodyGroupCount doesnt change

5911674

undo the minimal config change

f83b8af

Merge branch 'develop' into update-easlot-pruning

f2f5083

add context to CustodyGroupCount after merging from develop

98c164a

nalepae reviewed Oct 13, 2025

View reviewed changes

nalepae removed the optimization PR making Prysm more effective label Oct 13, 2025

satushh added 7 commits October 13, 2025 16:56

cosmetic change

cb7923a

shift responsibility from caller to callee, protection for updateEarl…

1143f0e

…iestSlot. UpdateEarliestAvailableSlot returns cgc

allow increase in earliestAvailableSlot only when custodyGroupCount a…

194b8a8

…lso increases

remove CustodyGroupCount as it is no longer needed as UpdateEarliestA…

e56094f

…vailableSlot returns cgc now

proper place for log and name refactor

2cab4d2

Merge branch 'develop' into update-easlot-pruning

7447610

Merge branch 'develop' into update-easlot-pruning

4552e18

nalepae reviewed Oct 16, 2025

View reviewed changes

test for Nil custody info

8ce6865

nalepae reviewed Oct 17, 2025

View reviewed changes

satushh added 4 commits October 17, 2025 11:51

allow decreasing earliest slot in DB (just like in memory)

b63f6bb

invert if statement to make more readable

17b67c2

UpdateEarliestAvailableSlot for DB (equivalent of p2p's UpdateEarlies…

8bd84a8

…tAvailableSlot) & undo changes made to UpdateCustodyInfo

Merge branch 'develop' into update-easlot-pruning

6a85a1c

	return nil
	return errors.Wrap(err, "failed to get custody group count for earliest available slot update")

		@@ -0,0 +1,3 @@
		### Added

		- Update the earliest available slot while pruning. No newline at end of file

Update Earliest available slot when pruning #15694

Are you sure you want to change the base?

Update Earliest available slot when pruning #15694

Uh oh!

Conversation

satushh commented Sep 15, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull Request Overview

Reviewed Changes

Uh oh!

Copilot AI Sep 16, 2025

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Copilot AI Sep 16, 2025

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Copilot AI Sep 16, 2025

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!

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!

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!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

satushh commented Sep 15, 2025 •

edited

Loading