ChainDB: let the BlockFetch client add blocks asynchronously #2721
+120
−42
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
|
Draft because the impact on bulk sync speed should be measured. |
mrBliss
force-pushed
the
mrBliss/add-blocks-fully-async
branch
2 times, most recently
from
6e98685
to
aa1413e
Compare
mrBliss
commented
Nov 3, 2020
| return $ \pt -> | ||
| case pointToWithOriginRealPoint pt of | ||
| Origin -> False | ||
| NotOrigin pt' -> checkBlocksToAdd pt' || checkVolDb (realPointHash pt') |
There was a problem hiding this comment.
Race condition: when a block has been removed but not yet written to the VolatileDB, this will return False. Possible solution: in the background thread, peek the block, process it, and then remove it from the queue. Add a comment that these two are not disjoint.
cc: @edsko.
There was a problem hiding this comment.
Good thing someone is paying attention 🙄
mrBliss
force-pushed
the
mrBliss/add-blocks-fully-async
branch
from
c5b3a28
to
2defb1f
Compare
Fixes #2487. Currently, the effective queue size when adding blocks to the ChainDB is 1 (for why, see #2487). In this commit, we let the BlockFetch client add blocks fully asynchronously to the ChainDB, which restores the effective queue size to the configured value again, e.g., 10. The BlockFetch client will no longer wait until the block has been written to the VolatileDB (and thus also not until the block has been processed by chain selection). The BlockFetch client can just hand over the block and continue downloading with minimum delay. To make this possible, we change the behaviour of `getIsFetched` and `getMaxSlotNo` to account for the blocks in the queue, otherwise the BlockFetch client might try to redownload already-fetched blocks. This is an alternative to #2489, which let the BlockFetch client write blocks to the VolatileDB synchronously. The problem with that approach is that multiple threads are writing to the VolatileDB, instead of a single background thread. We have relied on the latter to simplify the VolatileDB w.r.t. consistency after incomplete writes.
See the comment on `cdbBlocksToAdd`.
mrBliss
force-pushed
the
mrBliss/add-blocks-fully-async
branch
from
2defb1f
to
5299069
Compare
Niols
added a commit
to IntersectMBO/ouroboros-consensus
that referenced
this pull request
Jul 11, 2024
This is a port of PR IntersectMBO/ouroboros-network#2721 to the new ChainSelQueue. Co-authored-by: mrBliss <thomas@well-typed.com>
Niols
added a commit
to IntersectMBO/ouroboros-consensus
that referenced
this pull request
Jul 11, 2024
This is a port of PR IntersectMBO/ouroboros-network#2721 to the new ChainSelQueue. Co-authored-by: mrBliss <thomas@well-typed.com>
facundominguez
pushed a commit
to IntersectMBO/ouroboros-consensus
that referenced
this pull request
Jul 11, 2024
This is a port of PR IntersectMBO/ouroboros-network#2721 to the new ChainSelQueue. Co-authored-by: mrBliss <thomas@well-typed.com>
facundominguez
pushed a commit
to IntersectMBO/ouroboros-consensus
that referenced
this pull request
Jul 23, 2024
This is a port of PR IntersectMBO/ouroboros-network#2721 to the new ChainSelQueue. Co-authored-by: mrBliss <thomas@well-typed.com>
facundominguez
added a commit
to IntersectMBO/ouroboros-consensus
that referenced
this pull request
Aug 6, 2024
Port of IntersectMBO/ouroboros-network#2721 Co-Authored-By: Thomas Winant <thomas@well-typed.com>
amesgen
pushed a commit
to IntersectMBO/ouroboros-consensus
that referenced
this pull request
Aug 7, 2024
Port of IntersectMBO/ouroboros-network#2721 Co-Authored-By: Thomas Winant <thomas@well-typed.com>
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.
Fixes IntersectMBO/ouroboros-consensus#655.
Currently, the effective queue size when adding blocks to the ChainDB is 1 (for
why, see IntersectMBO/ouroboros-consensus#655). In this commit, we let the BlockFetch client add blocks fully
asynchronously to the ChainDB, which restores the effective queue size to the
configured value again, e.g., 10.
The BlockFetch client will no longer wait until the block has been written to
the VolatileDB (and thus also not until the block has been processed by chain
selection). The BlockFetch client can just hand over the block and continue
downloading with minimum delay. To make this possible, we change the behaviour
of
getIsFetchedandgetMaxSlotNoto account for the blocks in the queue,otherwise the BlockFetch client might try to redownload already-fetched blocks.
This is an alternative to #2489, which let the BlockFetch client write blocks to
the VolatileDB synchronously. The problem with that approach is that multiple
threads are writing to the VolatileDB, instead of a single background thread. We
have relied on the latter to simplify the VolatileDB w.r.t. consistency after
incomplete writes.