PHOENIX-7878 CDC perf improvement - skip redundant cell versions on data table scans#2493
Open
virajjasani wants to merge 2 commits into
Open
PHOENIX-7878 CDC perf improvement - skip redundant cell versions on data table scans#2493virajjasani wants to merge 2 commits into
virajjasani wants to merge 2 commits into
Conversation
Contributor
Author
palashc
approved these changes
Jun 3, 2026
| String cdcFullName = SchemaUtil.getTableName(schemaName, cdcName); | ||
| try (Connection conn = newConnection(tenantId)) { | ||
| // For debug: uncomment to see the exact results logged to console. | ||
| dumpCDCResults(conn, cdcName, new TreeMap<String, String>() { |
Contributor
There was a problem hiding this comment.
Was this meant to be commented out?
| } | ||
| return CDCUtil.setupScanForCDC(dataScan); | ||
| CDCUtil.setupScanForCDC(dataScan); | ||
| Map<ImmutableBytesPtr, long[]> timestampMap = buildDataRowTimestampMap(dataRowKeys); |
Contributor
There was a problem hiding this comment.
Can we avoid building this timestamp map in every task and precompute beforehand? But maybe it is okay since number of tasks will usually be small - based on number of regions involved and number of rowkeys?
Contributor
Author
There was a problem hiding this comment.
It is possible but it is far bigger refactor, worth doing as separate PR because this PR is already too big. However, dataRowKeys are already available so generating map would not be that time consuming.
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
2 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Jira: PHOENIX-7878
What changes were proposed in this pull request?
CDC perf improvement - skip redundant cell versions on data table scans
Why are the changes needed?
When a CDC query runs with pre, post, and/or change scopes, it scans the data table to reconstruct each change event (the change image plus the pre-image, and for the consumer path the full data-row state). Today that data table scan is a raw, all-versions scan, so for every data row we read back every version of every column - even though, for a given batch of changes, we only need two cells per column per change: the cell at the change timestamp, and the most recent cell just below it (the pre-image). On rows that are updated frequently this means we read, transfer, and process far more cells than the event reconstruction actually uses, which adds CPU, memory, and network overhead to CDC reads.
The purpose of this Jira is to add new CDCVersionFilter, in addition to SkipScanFilter on the data table scans. For each row it is given the set of change timestamps from the current batch and keeps only the cells that matter: the cell at each change timestamp, the first cell below each change timestamp (the pre-image), and all DeleteFamily markers (needed for deletion tracking), other cells are skipped to avoid redundant data transfer.
Does this PR introduce any user-facing change?
This is performance improvement
How was this patch tested?
UT and IT tests
Was this patch authored or co-authored using generative AI tooling?
Claude Opus 4.8