{{ message }}
Repository navigation
Conversation
debug_checkpoint_scan_sleep_ms makes a checkpoint sleep after it has written a table's row groups and before it installs them, so that tests can commit alongside a checkpoint at a known point.
… them This is a 1.5 backport for duckdb#25970 When a table's metadata is reused because none of its row groups changed, the checkpoint called CheckpointDeletes on every row group only to mark the row group's delete blocks as still in use. That serializes the version info if a delete committed after the row group was found unchanged, which is possible because commits do not wait for a checkpoint: it writes new delete blocks with the same content (the commit lies above the checkpoint's visibility bound), discards them, and leaves the old blocks, which the reused metadata still references, without the in-use mark. The checkpoint frees them, and once anything reuses them the table cannot be loaded anymore: replaying the WAL fails with "Pinning block 315.39 but it is marked as a free block". The unchanged path now only marks the persisted delete blocks in use (RowGroup::GetPersistedDeletePointers) and never serializes; the delete that committed meanwhile is written by the next checkpoint. That is also why RowVersionManager::GetStoragePointers no longer asserts that no delete commit is pending: the blocks it returns are the ones the last checkpoint wrote, a pending commit changes nothing until the next one. This is the minimal fix for the release branch. duckdb#25970 fixes the same problem on v2.0 in the version manager itself, by letting a checkpoint ignore deletes that committed above its bound.
duckdb-draftbot
marked this pull request as draft
September 21, 2026 08:34
ywelsch
marked this pull request as ready for review
September 21, 2026 08:37
Member
This file contains hidden or 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
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.
This is a 1.5 backport for #25970
When a table's metadata is reused because none of its row groups changed, the checkpoint called CheckpointDeletes on every row group only to mark the row group's delete blocks as still in use. That serializes the version info if a delete committed after the row group was found unchanged, which is possible because commits do not wait for a checkpoint: it writes new delete blocks with the same content (the commit lies above the checkpoint's visibility bound), discards them, and leaves the old blocks, which the reused metadata still references, without the in-use mark. The checkpoint frees them, and once anything reuses them the table cannot be loaded anymore: replaying the WAL fails with
"Pinning block 315.39 but it is marked as a free block".The unchanged path now only marks the persisted delete blocks in use (RowGroup::GetPersistedDeletePointers) and never serializes; the delete that committed meanwhile is written by the next checkpoint. That is also why RowVersionManager::GetStoragePointers no longer asserts that no delete commit is pending: the blocks it returns are the ones the last checkpoint wrote, a pending commit changes nothing until the next one.
This is the minimal fix for the release branch. #25970 fixes the same problem on v2.0 in the version manager itself, by letting a checkpoint ignore deletes that committed above its bound.