storage: String interpolation - #37903
Draft
peterdukelarsen wants to merge 3 commits into
Draft
Conversation
Flip the mysql_source_snapshot_parallelism dyncfg default from true to false, so new deployments use the legacy single-worker-per-table snapshot unless the flag is enabled. Parallel PK-range snapshotting stays available behind the flag and is now rolled out selectively via LaunchDarkly rather than being on everywhere by default. CI continues to run with the flag enabled through the mzcompose default system parameters, so the parallel path keeps its test coverage.
Replace the OFFSET-walk boundary sampling for parallel MySQL snapshots with interpolation between MIN and MAX of the primary key. The old walk cost a full index pass per table up front. The new computation is two index dives plus integer math, at the price of scoping support down to single-column string-like primary keys with a binary collation (ascii_bin, latin1_bin, utf8mb4_bin, utf8mb4_0900_bin), where MySQL's comparison order matches the digit order the math preserves. Integer primary keys now take the single-worker fallback. Keys are mapped to digit strings (bytes for single-byte charsets, code points with the surrogate gap collapsed for utf8mb4), interpolated as fixed-precision integers after stripping the shared prefix, and shipped as hex literals with a charset introducer. All boundaries share one digit length, so their order is unaffected by PAD SPACE vs NO PAD. The leader additionally verifies boundary order with MySQL before broadcasting and falls back if the check fails, and the existing worker-side re-validation still guards against collation-change races. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Before counting a table for the snapshot size gauge, read the optimizer's row estimate from information_schema.tables. Tables estimated above mysql_source_snapshot_exact_count_max_rows (default 1M) report the estimate directly instead of paying an O(rows) COUNT(*) index walk. The gauge only drives progress reporting, so an estimate is a fair trade on large tables, and small tables keep exact counts so statistics tests stay deterministic. The threshold is a dyncfg so tests can force the estimated path: mysql-cdc/statistics.td lowers it below a 1000-row table and asserts the gauge lands in a sane band while the staged count stays exact, and parallel-workload flips it between 0 and the default. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
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
1 participant
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.
Motivation
Computing partition boundaries by scanning the table in MySQL is expensive enough that we aren't seeing improvements in snapshot performance. Instead, we want to compute cheap partition boundaries.
Description
Verification
How do you know this change is correct? Describe new or existing automated
tests, or manual steps you took.