ndimiduk commented on code in PR #5170:
URL: https://github.com/apache/hbase/pull/5170#discussion_r1168471284
##########
hbase-server/src/main/java/org/apache/hadoop/hbase/master/normalizer/SimpleRegionNormalizer.java:
##########
@@ -474,6 +494,11 @@ private boolean isLargeEnoughForMerge(final
NormalizerConfiguration normalizerCo
return getRegionSizeMB(regionInfo) >=
normalizerConfiguration.getMergeMinRegionSizeMb(ctx);
}
+ private boolean isPresplitEmptyRegion(final NormalizerConfiguration
normalizerConfiguration,
+ final NormalizeContext ctx, final RegionInfo regionInfo) {
+ return getStoreFileCount(regionInfo) == 0 &&
Review Comment:
I think that this single empty store file behaviour is a coincidence, not
something that we want to design a functionality around.
Original implementation of normaliser had a long discussion about how to
handle pre-split regions. As I recall, the conclusion was to use the minimum
age setting as a means for an operator to preserve an empty, pre-split table. I
don't particularly like this approach either, but at least it means that any
forgotten empty regions will eventually converge on a more correct state.
If you want to pursue some other means, I prefer it involve explicit
metadata instead of more implicit assumptions (especially about implementation
details). Sorry @chrajeshbabu , but I strongly disapprove of this approach.
--
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]
