[Merged by Bors] - fix deneb sync bug #4869
Closed
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
bors bot
pushed a commit
that referenced
this pull request
Oct 25, 2023
1. Add commitments to logs and update the `Display` implementation of `KzgCommitment` to become truncated similarly to block root.
I've been finding it difficult to debug scenarios involving multiple blobs for the same `(index, block_root)`. Logging the commitment will help with this, we can match it to what exists in the block.
Example output:
```
Oct 20 21:13:36.700 DEBG Successfully verified gossip blob commitment: 0xa3c1…1cd8, index: 0, root: 0xf31e…f9de, slot: 154568
Oct 20 21:13:36.785 DEBG Successfully verified gossip block commitments: [0xa3c1…1cd8, 0x8655…02ff, 0x8d6a…955a, 0x84ac…3a1b, 0x9752…629b, 0xb9fc…20fb], root: 0xf31eeb732702e429e89057b15e1c0c631e8452e09e03cb1924353f536ef4f9de, slot: 154568, graffiti: teku/besu, service: beacon
```
Example output in a block with no blobs (this will show up pre-deneb):
```
426734:Oct 20 21:15:24.113 DEBG Successfully verified gossip block, commitments: [], root: 0x619db1360ba0e8d44ae2a0f2450ebca47e167191feecffcfac0e8d7b6c39623c, slot: 154577, graffiti: teku/nethermind, service: beacon, module: beacon_chain::beacon_chain:2765
```
2. Remove `strum::IntoStaticStr` from `AvailabilityCheckError`. This is because `IntoStaticStr` end up dropping information inside the enum. So kzg commitments in this error are dropped, making it more difficult to debug
```
AvailabilityCheckError::KzgCommitmentMismatch {
blob_commitment: KzgCommitment,
block_commitment: KzgCommitment,
},
```
which is output as just `AvailabilityCheckError`
3. Some additional misc sync logs I found useful in debugging #4869
4. This downgrades ”Block returned for single block lookup not present” to debug because I don’t think we can fix the scenario that causes this unless we can cancel inflight rpc requests
Co-authored-by: realbigsean <seananderson33@gmail.com>
bors bot
pushed a commit
that referenced
this pull request
Oct 25, 2023
1. Add commitments to logs and update the `Display` implementation of `KzgCommitment` to become truncated similarly to block root.
I've been finding it difficult to debug scenarios involving multiple blobs for the same `(index, block_root)`. Logging the commitment will help with this, we can match it to what exists in the block.
Example output:
```
Oct 20 21:13:36.700 DEBG Successfully verified gossip blob commitment: 0xa3c1…1cd8, index: 0, root: 0xf31e…f9de, slot: 154568
Oct 20 21:13:36.785 DEBG Successfully verified gossip block commitments: [0xa3c1…1cd8, 0x8655…02ff, 0x8d6a…955a, 0x84ac…3a1b, 0x9752…629b, 0xb9fc…20fb], root: 0xf31eeb732702e429e89057b15e1c0c631e8452e09e03cb1924353f536ef4f9de, slot: 154568, graffiti: teku/besu, service: beacon
```
Example output in a block with no blobs (this will show up pre-deneb):
```
426734:Oct 20 21:15:24.113 DEBG Successfully verified gossip block, commitments: [], root: 0x619db1360ba0e8d44ae2a0f2450ebca47e167191feecffcfac0e8d7b6c39623c, slot: 154577, graffiti: teku/nethermind, service: beacon, module: beacon_chain::beacon_chain:2765
```
2. Remove `strum::IntoStaticStr` from `AvailabilityCheckError`. This is because `IntoStaticStr` end up dropping information inside the enum. So kzg commitments in this error are dropped, making it more difficult to debug
```
AvailabilityCheckError::KzgCommitmentMismatch {
blob_commitment: KzgCommitment,
block_commitment: KzgCommitment,
},
```
which is output as just `AvailabilityCheckError`
3. Some additional misc sync logs I found useful in debugging #4869
4. This downgrades ”Block returned for single block lookup not present” to debug because I don’t think we can fix the scenario that causes this unless we can cancel inflight rpc requests
Co-authored-by: realbigsean <seananderson33@gmail.com>
ethDreamer
approved these changes
Oct 31, 2023
There was a problem hiding this comment.
LGTM! Not that I understand sync all that well but this seems to make sense. Lighthouse was continuously stalling yesterday when syncing devnet-10 and then pawan pointed me to this PR. Unfortunately devnet-10 was shut down by the time I tried this PR this morning but lighthouse does seem to sync fine on devnet-11. This devnet is much newer though so we might have to wait a bit before we can hit this bug again on devnet-11 and retry..
|
bors r+ |
bors bot
pushed a commit
that referenced
this pull request
Oct 31, 2023
## Issue Addressed I observed our forward sync on devnet 9 would stall when we would hit this log: ``` 250425:Oct 19 00:54:17.133 WARN Blocks and blobs request for range received invalid data, error: KzgCommitmentMismatch, batch_id: 4338, peer_id: 16Uiu2HAmHbmkEQFDrJfNuy1aYyAfHkNUwSD9FN7EVAqGJ8YTF9Mh, service: sync, module: network::sync::manager:1036 ``` ## Proposed Changes `range_sync_block_and_blob_response` [here](https://github.com/sigp/lighthouse/blob/1cb02a13a53d0e603ad5920c03832e5779c3df61/beacon_node/network/src/sync/manager.rs#L1013) removes the request from the sync manager. later, however if there's an error, `inject_error` [here](https://github.com/sigp/lighthouse/blob/1cb02a13a53d0e603ad5920c03832e5779c3df61/beacon_node/network/src/sync/manager.rs#L1055) expects the request to exist so we can handle retry logic. So this PR just re-inserts the request (withthout any accumulated blobs or blocks) when we hit an error here. The issue is unique to block+blob sync because the error here is only possible from mismatches between blocks + blobs after we've downloaded both, there's no equivalent error in block sync Co-authored-by: realbigsean <seananderson33@gmail.com>
|
Build failed: |
|
bors retry |
bors bot
pushed a commit
that referenced
this pull request
Oct 31, 2023
## Issue Addressed I observed our forward sync on devnet 9 would stall when we would hit this log: ``` 250425:Oct 19 00:54:17.133 WARN Blocks and blobs request for range received invalid data, error: KzgCommitmentMismatch, batch_id: 4338, peer_id: 16Uiu2HAmHbmkEQFDrJfNuy1aYyAfHkNUwSD9FN7EVAqGJ8YTF9Mh, service: sync, module: network::sync::manager:1036 ``` ## Proposed Changes `range_sync_block_and_blob_response` [here](https://github.com/sigp/lighthouse/blob/1cb02a13a53d0e603ad5920c03832e5779c3df61/beacon_node/network/src/sync/manager.rs#L1013) removes the request from the sync manager. later, however if there's an error, `inject_error` [here](https://github.com/sigp/lighthouse/blob/1cb02a13a53d0e603ad5920c03832e5779c3df61/beacon_node/network/src/sync/manager.rs#L1055) expects the request to exist so we can handle retry logic. So this PR just re-inserts the request (withthout any accumulated blobs or blocks) when we hit an error here. The issue is unique to block+blob sync because the error here is only possible from mismatches between blocks + blobs after we've downloaded both, there's no equivalent error in block sync Co-authored-by: realbigsean <seananderson33@gmail.com>
|
Pull request successfully merged into unstable. Build succeeded! The publicly hosted instance of bors-ng is deprecated and will go away soon. If you want to self-host your own instance, instructions are here. If you want to switch to GitHub's built-in merge queue, visit their help page.
|
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Issue Addressed
I observed our forward sync on devnet 9 would stall when we would hit this log:
Proposed Changes
range_sync_block_and_blob_responsehere removes the request from the sync manager. later, however if there's an error,inject_errorhere expects the request to exist so we can handle retry logic. So this PR just re-inserts the request (withthout any accumulated blobs or blocks) when we hit an error here.The issue is unique to block+blob sync because the error here is only possible from mismatches between blocks + blobs after we've downloaded both, there's no equivalent error in block sync