SaadASTheDev opened a new pull request, #8498:
URL: https://github.com/apache/hbase/pull/8498
## Summary
`SimpleRegionNormalizer` incorrectly includes secondary replica regions when
computing normalization plans. Since `getRegionsOfTable` returns all replicas
alongside their primaries, and `RegionInfo.COMPARATOR` sorts replicas
immediately after their primary (same start/end key, higher replica ID), the
merge planning loop treats primaries and replicas as contiguous candidates
and
groups them together into a single `MergeNormalizationPlan`.
These plans are rejected at the `MergeTableRegionsProcedure` level with:
MergeRegionException: Can't merge non-default replicas
This causes repeated failed plan submissions, wasted work, and can prevent
normalization from making progress when region replication is enabled.
## Root Cause
`NormalizeContext` builds its region list from `getRegionsOfTable`, which
returns all replicas. The normalizer has no concept of replica IDs and treats
every entry as a candidate for merge or split.
## Fix
Filter non-default replicas out of `tableRegions` in `NormalizeContext`
immediately after fetching, before sorting and before computing the average
region size. Only primary regions (replicaId == DEFAULT_REPLICA_ID == 0)
are valid normalization targets. When a primary is merged or split,
`MergeTableRegionsProcedure` and `SplitTableRegionProcedure` automatically
handle all associated replicas by reading the table's region replication
factor and expanding the operation to cover every replica ID.
## Tests
Added `itIgnoresSecondaryReplicasForMergeAndSplitPlanning` to
`TestSimpleRegionNormalizer`, which exercises the normalizer with a mixed
list of primary and replica regions and asserts that only primary regions
appear in the resulting normalization plans.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
