Truthound · Storage and recovery consistency · 2/3
Check rows, values and metadata before and after storage.
55 articles
Keep Hudi record-index scope unchanged during a size rebuild
Check that a size-oriented index rebuild preserved its partitioned or global scope. Reconcile the prior definition with effective configuration rather than inferring identity from a shared metadata name.
Match Hudi incremental pruning statistics to historical file coverage
A current snapshot can conceal pruning omissions in a historical incremental read. Fix the requested timeline range and reconcile its keys with a control read covering that same range.
Apply the version correction before testing string-coordinate storage
Do not generalize an old-version failure to current support. Admit string coordinates using a roundtrip on the actual installation.
Check the returned drop_encoding object when NetCDF coordinates truncate
Approve in-memory concatenation and final-file coordinates separately. Do not reject a workaround tested without retaining its returned object.
Verify the UTC instant after a future DST roundtrip
A retained timezone name does not establish the same instant. Keep the local representation beside the epoch value.
Why unknown NaN-page statistics cannot be omitted from row-group bounds
An unknown range is different from an empty page. Check that the advertised lower bound does not exclude original valid values.
Do not read truncated Delta timestamp statistics as exact upper bounds
Preserve the precision difference between millisecond statistics and microsecond values. Read duplicate reporting and the follow-up fix together.
Restore expression objects before accepting a DuckDB export as recoverable
Restore schemas containing COLLATE and subscripts into a separate database. Distinguishing export completion, SQL parse success, and restored values identifies the first failing recovery boundary.
Wait for Doris load visibility across a schema-change barrier
Link COMMITTED loads to visible keys before announcing completion during a schema change. Investigate transaction and version traces instead of treating a visibility timeout as an automatic reason to reload.
Check the logical slice boundary before subtracting a Doris read offset
Check the logical slice boundary before relying on the enclosing packed object's size. Separate direct-reader and cache-protected paths when investigating an out-of-range offset.
Use one metadata snapshot before deleting a Doris packed slice
Build packed-slice deletion candidates from every page of formal, recycle, and temporary references at one snapshot. Absence in inventories captured at different times does not establish absence overall.
Include newly cloned Doris replicas in a binlog TTL transition
After extending binlog retention, check whether replicas cloned during the transition received the new TTL. Record the policy generation alongside current serving membership to find replicas outside the original completion responses.
Inspect nullable and comment metadata when Delta inventory schemas are rejected
Investigate a VACUUM inventory rejection by separating dtype mismatches from field metadata differences such as nullability and comments. Preserve the original schema and keep DRY RUN review separate from deletion approval.
Avoid counting a Delta AddFile re-add as storage growth
For metadata-only re-adds, reconcile CRC values against active state that preserves file identity rather than counting actions.
Check Delta checkpoint actions even when Parquet parses
A readable Parquet file can still be the output of a failed checkpoint write. Check actual protocol and metaData actions against the source log range before approving the candidate.
Preserve untargeted Nested siblings during ClickHouse TTL mutation
After a Nested-field TTL, reconcile sibling arrays outside the intended mutation scope. Check that required business values survived separately from CHECK TABLE success before allowing further mutations.
Trino filters after a Hive decimal change: inspect Parquet storage type and scale
Interpreting old integer-file statistics through only the current schema can prune matching rows.
Rows visible in Trino after Hive ACID TRUNCATE: validate the transaction state
The consumer's approved post-TRUNCATE view matters more than whether old physical files still exist.
RisingWave DELETE after an upsert: inspect prior state before cancelling operations
Deleting a new insertion and deleting an overwrite of an existing row are different state transitions.
Polars zero-column round trips: preserve height 3 as independent state
An empty column array does not imply zero rows. Check explicit height across an extension interface.
