[
https://issues.apache.org/jira/browse/HBASE-14491?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14955440#comment-14955440
]
Enis Soztutar commented on HBASE-14491:
---------------------------------------
In HRegion.doMiniBatchMutation(), we are preparing the WALEdits, by appending
cells to the WALEdits one by by caling {{WALEdit.add(Cell)}}.
{code}
Map<byte[], List<Cell>>[] familyMaps = new Map[batchOp.operations.length];
...
Map<byte[], List<Cell>> familyMap = mutation.getFamilyCellMap();
...
private void addFamilyMapToWALEdit(Map<byte[], List<Cell>> familyMap,
WALEdit walEdit) {
for (List<Cell> edits : familyMap.values()) {
assert edits instanceof RandomAccess;
int listSize = edits.size();
for (int i=0; i < listSize; i++) {
Cell cell = edits.get(i);
walEdit.add(cell);
}
}
}
{code}
My understanding is that the cells within a WALEdit will be grouped by Mutation
which is a single row. There can be multiple mutations in the same WALEdit
sharing the same row key, but they are still distinct mutations that has to be
counted separately. I would say we should resolve this jira now.
> ReplicationSource#countDistinctRowKeys code logic is not correct
> ----------------------------------------------------------------
>
> Key: HBASE-14491
> URL: https://issues.apache.org/jira/browse/HBASE-14491
> Project: HBase
> Issue Type: Bug
> Reporter: Ashish Singhi
> Assignee: Ashish Singhi
> Priority: Minor
>
> {code}
> Cell lastCell = cells.get(0);
> for (int i = 0; i < edit.size(); i++) {
> if (!CellUtil.matchingRow(cells.get(i), lastCell)) {
> distinctRowKeys++;
> }
> }
> {code}
> The above logic for finding the distinct row keys in the list needs to be
> corrected.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
