PeerDAS: Multiple improvements #14467
Merged
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
nalepae
requested review from
kasey,
terencechain and
james-prysm
and removed request for
a team
… some columns. Before this commit: Each time we receive a column, we look into the filesystem for all columns we store. ==> For 128 columns, it looks for 1 + 2 + 3 + ... + 128 = 128(128+1)/2 = 8256 files look. Also, as soon as a column is saved into the file system, then if, right after, we look at the filesystem again, we assume the column will be available (strict consistency). It happens not to be always true. ==> Sometimes, we can reconstruct and reseed columns more than once, because of this lack of filesystem strict consistency. After this commit: We use a (strictly consistent) cache to determine if we received a column or not. ==> No more consistency issue, and less stress for the filesystem.
Before this commit, logged values assumed that all requested columns correspond to the same block root, which is not always the case. After this commit, we know which columns are requested for which root.
This is useful to debug "lost data columns" in devnet.
nalepae
force-pushed
the
peerdas-misc
branch
from
40fdb9e
to
de90f26
Compare
nisdas
reviewed
Sep 23, 2024
beacon-chain/p2p/broadcaster.go
Outdated
|
|
||
| return nil | ||
| } | ||
|
|
||
| func (s *Service) internalBroadcastDataColumn( | ||
| ctx context.Context, | ||
| slot primitives.Slot, |
There was a problem hiding this comment.
Do we need the slot and slotStartTime if we can already get this information from the sidecar object
There was a problem hiding this comment.
Addressed here: 1dc8225
Additional question:
We also have the possibility to remove the root argument from internalBroadcasDatacolumn, since the hash tree root can be obtain from the data columns sidecar object with:
dataColumnSidecar.SignedBlockHeader.GetHeader().HashTreeRoot()However, is the cost of re-computing HashTreeRoot worth the removal of the root argument in the function?
There was a problem hiding this comment.
I think passing the root is a better idea, hashing the header 128 times will waste CPU cycles unnecessarily.
| dataColumnsWithholdCount := features.Get().DataColumnsWithholdCount | ||
|
|
||
| // Get the time corresponding to the start of the slot. | ||
| genesisTime := uint64(vs.TimeFetcher.GenesisTime().Unix()) | ||
| slotStartTime, err := slots.ToTime(genesisTime, slot) |
There was a problem hiding this comment.
As mentioned earlier , I think this can be done in the broadcast method itself, we do not need to provide it here.
There was a problem hiding this comment.
Addressed in the other comment.
nisdas
approved these changes
Sep 23, 2024
The context is explained here. |
nalepae
added a commit
that referenced
this pull request
Oct 7, 2024
* `scheduleReconstructedDataColumnsBroadcast`: Really minor refactor. * `receivedDataColumnsFromRootLock` -> `dataColumnsFromRootLock` * `reconstructDataColumns`: Stop looking into the DB to know if we have some columns. Before this commit: Each time we receive a column, we look into the filesystem for all columns we store. ==> For 128 columns, it looks for 1 + 2 + 3 + ... + 128 = 128(128+1)/2 = 8256 files look. Also, as soon as a column is saved into the file system, then if, right after, we look at the filesystem again, we assume the column will be available (strict consistency). It happens not to be always true. ==> Sometimes, we can reconstruct and reseed columns more than once, because of this lack of filesystem strict consistency. After this commit: We use a (strictly consistent) cache to determine if we received a column or not. ==> No more consistency issue, and less stress for the filesystem. * `dataColumnSidecarByRootRPCHandler`: Improve logging. Before this commit, logged values assumed that all requested columns correspond to the same block root, which is not always the case. After this commit, we know which columns are requested for which root. * Add a log when broadcasting a data column. This is useful to debug "lost data columns" in devnet. * Address Nishant's comment
nalepae
added a commit
that referenced
this pull request
Oct 7, 2024
* `scheduleReconstructedDataColumnsBroadcast`: Really minor refactor. * `receivedDataColumnsFromRootLock` -> `dataColumnsFromRootLock` * `reconstructDataColumns`: Stop looking into the DB to know if we have some columns. Before this commit: Each time we receive a column, we look into the filesystem for all columns we store. ==> For 128 columns, it looks for 1 + 2 + 3 + ... + 128 = 128(128+1)/2 = 8256 files look. Also, as soon as a column is saved into the file system, then if, right after, we look at the filesystem again, we assume the column will be available (strict consistency). It happens not to be always true. ==> Sometimes, we can reconstruct and reseed columns more than once, because of this lack of filesystem strict consistency. After this commit: We use a (strictly consistent) cache to determine if we received a column or not. ==> No more consistency issue, and less stress for the filesystem. * `dataColumnSidecarByRootRPCHandler`: Improve logging. Before this commit, logged values assumed that all requested columns correspond to the same block root, which is not always the case. After this commit, we know which columns are requested for which root. * Add a log when broadcasting a data column. This is useful to debug "lost data columns" in devnet. * Address Nishant's comment
nalepae
added a commit
that referenced
this pull request
Oct 7, 2024
* `scheduleReconstructedDataColumnsBroadcast`: Really minor refactor. * `receivedDataColumnsFromRootLock` -> `dataColumnsFromRootLock` * `reconstructDataColumns`: Stop looking into the DB to know if we have some columns. Before this commit: Each time we receive a column, we look into the filesystem for all columns we store. ==> For 128 columns, it looks for 1 + 2 + 3 + ... + 128 = 128(128+1)/2 = 8256 files look. Also, as soon as a column is saved into the file system, then if, right after, we look at the filesystem again, we assume the column will be available (strict consistency). It happens not to be always true. ==> Sometimes, we can reconstruct and reseed columns more than once, because of this lack of filesystem strict consistency. After this commit: We use a (strictly consistent) cache to determine if we received a column or not. ==> No more consistency issue, and less stress for the filesystem. * `dataColumnSidecarByRootRPCHandler`: Improve logging. Before this commit, logged values assumed that all requested columns correspond to the same block root, which is not always the case. After this commit, we know which columns are requested for which root. * Add a log when broadcasting a data column. This is useful to debug "lost data columns" in devnet. * Address Nishant's comment
nalepae
added a commit
that referenced
this pull request
Oct 7, 2024
* `scheduleReconstructedDataColumnsBroadcast`: Really minor refactor. * `receivedDataColumnsFromRootLock` -> `dataColumnsFromRootLock` * `reconstructDataColumns`: Stop looking into the DB to know if we have some columns. Before this commit: Each time we receive a column, we look into the filesystem for all columns we store. ==> For 128 columns, it looks for 1 + 2 + 3 + ... + 128 = 128(128+1)/2 = 8256 files look. Also, as soon as a column is saved into the file system, then if, right after, we look at the filesystem again, we assume the column will be available (strict consistency). It happens not to be always true. ==> Sometimes, we can reconstruct and reseed columns more than once, because of this lack of filesystem strict consistency. After this commit: We use a (strictly consistent) cache to determine if we received a column or not. ==> No more consistency issue, and less stress for the filesystem. * `dataColumnSidecarByRootRPCHandler`: Improve logging. Before this commit, logged values assumed that all requested columns correspond to the same block root, which is not always the case. After this commit, we know which columns are requested for which root. * Add a log when broadcasting a data column. This is useful to debug "lost data columns" in devnet. * Address Nishant's comment
nalepae
added a commit
that referenced
this pull request
Oct 8, 2024
* `scheduleReconstructedDataColumnsBroadcast`: Really minor refactor. * `receivedDataColumnsFromRootLock` -> `dataColumnsFromRootLock` * `reconstructDataColumns`: Stop looking into the DB to know if we have some columns. Before this commit: Each time we receive a column, we look into the filesystem for all columns we store. ==> For 128 columns, it looks for 1 + 2 + 3 + ... + 128 = 128(128+1)/2 = 8256 files look. Also, as soon as a column is saved into the file system, then if, right after, we look at the filesystem again, we assume the column will be available (strict consistency). It happens not to be always true. ==> Sometimes, we can reconstruct and reseed columns more than once, because of this lack of filesystem strict consistency. After this commit: We use a (strictly consistent) cache to determine if we received a column or not. ==> No more consistency issue, and less stress for the filesystem. * `dataColumnSidecarByRootRPCHandler`: Improve logging. Before this commit, logged values assumed that all requested columns correspond to the same block root, which is not always the case. After this commit, we know which columns are requested for which root. * Add a log when broadcasting a data column. This is useful to debug "lost data columns" in devnet. * Address Nishant's comment
nalepae
added a commit
that referenced
this pull request
Oct 23, 2024
* `scheduleReconstructedDataColumnsBroadcast`: Really minor refactor. * `receivedDataColumnsFromRootLock` -> `dataColumnsFromRootLock` * `reconstructDataColumns`: Stop looking into the DB to know if we have some columns. Before this commit: Each time we receive a column, we look into the filesystem for all columns we store. ==> For 128 columns, it looks for 1 + 2 + 3 + ... + 128 = 128(128+1)/2 = 8256 files look. Also, as soon as a column is saved into the file system, then if, right after, we look at the filesystem again, we assume the column will be available (strict consistency). It happens not to be always true. ==> Sometimes, we can reconstruct and reseed columns more than once, because of this lack of filesystem strict consistency. After this commit: We use a (strictly consistent) cache to determine if we received a column or not. ==> No more consistency issue, and less stress for the filesystem. * `dataColumnSidecarByRootRPCHandler`: Improve logging. Before this commit, logged values assumed that all requested columns correspond to the same block root, which is not always the case. After this commit, we know which columns are requested for which root. * Add a log when broadcasting a data column. This is useful to debug "lost data columns" in devnet. * Address Nishant's comment
nalepae
added a commit
that referenced
this pull request
Oct 23, 2024
* `scheduleReconstructedDataColumnsBroadcast`: Really minor refactor. * `receivedDataColumnsFromRootLock` -> `dataColumnsFromRootLock` * `reconstructDataColumns`: Stop looking into the DB to know if we have some columns. Before this commit: Each time we receive a column, we look into the filesystem for all columns we store. ==> For 128 columns, it looks for 1 + 2 + 3 + ... + 128 = 128(128+1)/2 = 8256 files look. Also, as soon as a column is saved into the file system, then if, right after, we look at the filesystem again, we assume the column will be available (strict consistency). It happens not to be always true. ==> Sometimes, we can reconstruct and reseed columns more than once, because of this lack of filesystem strict consistency. After this commit: We use a (strictly consistent) cache to determine if we received a column or not. ==> No more consistency issue, and less stress for the filesystem. * `dataColumnSidecarByRootRPCHandler`: Improve logging. Before this commit, logged values assumed that all requested columns correspond to the same block root, which is not always the case. After this commit, we know which columns are requested for which root. * Add a log when broadcasting a data column. This is useful to debug "lost data columns" in devnet. * Address Nishant's comment
nalepae
added a commit
that referenced
this pull request
Oct 23, 2024
* `scheduleReconstructedDataColumnsBroadcast`: Really minor refactor. * `receivedDataColumnsFromRootLock` -> `dataColumnsFromRootLock` * `reconstructDataColumns`: Stop looking into the DB to know if we have some columns. Before this commit: Each time we receive a column, we look into the filesystem for all columns we store. ==> For 128 columns, it looks for 1 + 2 + 3 + ... + 128 = 128(128+1)/2 = 8256 files look. Also, as soon as a column is saved into the file system, then if, right after, we look at the filesystem again, we assume the column will be available (strict consistency). It happens not to be always true. ==> Sometimes, we can reconstruct and reseed columns more than once, because of this lack of filesystem strict consistency. After this commit: We use a (strictly consistent) cache to determine if we received a column or not. ==> No more consistency issue, and less stress for the filesystem. * `dataColumnSidecarByRootRPCHandler`: Improve logging. Before this commit, logged values assumed that all requested columns correspond to the same block root, which is not always the case. After this commit, we know which columns are requested for which root. * Add a log when broadcasting a data column. This is useful to debug "lost data columns" in devnet. * Address Nishant's comment
nalepae
added a commit
that referenced
this pull request
Oct 24, 2024
* `scheduleReconstructedDataColumnsBroadcast`: Really minor refactor. * `receivedDataColumnsFromRootLock` -> `dataColumnsFromRootLock` * `reconstructDataColumns`: Stop looking into the DB to know if we have some columns. Before this commit: Each time we receive a column, we look into the filesystem for all columns we store. ==> For 128 columns, it looks for 1 + 2 + 3 + ... + 128 = 128(128+1)/2 = 8256 files look. Also, as soon as a column is saved into the file system, then if, right after, we look at the filesystem again, we assume the column will be available (strict consistency). It happens not to be always true. ==> Sometimes, we can reconstruct and reseed columns more than once, because of this lack of filesystem strict consistency. After this commit: We use a (strictly consistent) cache to determine if we received a column or not. ==> No more consistency issue, and less stress for the filesystem. * `dataColumnSidecarByRootRPCHandler`: Improve logging. Before this commit, logged values assumed that all requested columns correspond to the same block root, which is not always the case. After this commit, we know which columns are requested for which root. * Add a log when broadcasting a data column. This is useful to debug "lost data columns" in devnet. * Address Nishant's comment
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.
Please read all the message commits containing needed information.