This is an automated email from the ASF dual-hosted git repository.
junegunn pushed a commit to branch master
in repository https://gitbox.apache.org/repos/asf/hbase.git
The following commit(s) were added to refs/heads/master by this push:
new 074024a226b HBASE-30351 ROWS_SCANNED scan metric double-counts rows
passing the filter when the scan uses the joined heap (#8596)
074024a226b is described below
commit 074024a226b78f2a4f00ac9ba44a7774a2f2ddc8
Author: eomiks <[email protected]>
AuthorDate: Tue Sep 1 19:50:00 2026 +0900
HBASE-30351 ROWS_SCANNED scan metric double-counts rows passing the filter
when the scan uses the joined heap (#8596)
Signed-off-by: Junegunn Choi <[email protected]>
---
.../hbase/regionserver/RegionScannerImpl.java | 5 ++-
.../TestServerSideScanMetricsFromClientSide.java | 51 ++++++++++++++++++++++
2 files changed, 55 insertions(+), 1 deletion(-)
diff --git
a/hbase-server/src/main/java/org/apache/hadoop/hbase/regionserver/RegionScannerImpl.java
b/hbase-server/src/main/java/org/apache/hadoop/hbase/regionserver/RegionScannerImpl.java
index 1c3530e7996..8a4e90429ad 100644
---
a/hbase-server/src/main/java/org/apache/hadoop/hbase/regionserver/RegionScannerImpl.java
+++
b/hbase-server/src/main/java/org/apache/hadoop/hbase/regionserver/RegionScannerImpl.java
@@ -359,7 +359,10 @@ public class RegionScannerImpl implements RegionScanner,
Shipper, RpcCallback {
nextKv = heap.peek();
moreCellsInRow = moreCellsInRow(nextKv, currentRowCell);
- if (!moreCellsInRow) {
+ // A row is scanned once its cells have been read from the store heap.
The joined heap only
+ // re-populates rows that already passed the filter with the cells of
the non essential
+ // families (HBASE-5416), so counting a completed row there would double
count it.
+ if (!moreCellsInRow && heap == this.storeHeap) {
incrementCountOfRowsScannedMetric(scannerContext);
}
if (moreCellsInRow && scannerContext.checkBatchLimit(limitScope)) {
diff --git
a/hbase-server/src/test/java/org/apache/hadoop/hbase/TestServerSideScanMetricsFromClientSide.java
b/hbase-server/src/test/java/org/apache/hadoop/hbase/TestServerSideScanMetricsFromClientSide.java
index 93b9c5d97b6..de3b48da7c8 100644
---
a/hbase-server/src/test/java/org/apache/hadoop/hbase/TestServerSideScanMetricsFromClientSide.java
+++
b/hbase-server/src/test/java/org/apache/hadoop/hbase/TestServerSideScanMetricsFromClientSide.java
@@ -245,6 +245,57 @@ public class TestServerSideScanMetricsFromClientSide {
testMetric(scan,
ServerSideScanMetrics.COUNT_OF_ROWS_SCANNED_KEY_METRIC_NAME, ROWS.length);
}
+ @Test
+ public void testRowsSeenMetricWithJoinedHeap() throws Exception {
+ // When Scan#setLoadColumnFamiliesOnDemand is enabled and the filter
declares some families
+ // as non essential, a row passing the filter is populated in two steps:
the essential
+ // families through the store heap and the remaining ones through the
joined heap
+ // (HBASE-5416). Such a row must still be counted only once in the
ROWS_SCANNED metric.
+ TableName tableName =
TableName.valueOf("testRowsSeenMetricWithJoinedHeap");
+ byte[] essentialFamily = Bytes.toBytes("essential");
+ byte[] joinedFamily = Bytes.toBytes("joined");
+ byte[] otherValue = Bytes.toBytes("otherValue");
+ int numMatchingRows = 5;
+ try (Table table =
+ TEST_UTIL.createTable(tableName, new byte[][] { essentialFamily,
joinedFamily })) {
+ List<Put> puts = new ArrayList<>();
+ for (int row = 0; row < NUM_ROWS; row++) {
+ Put put = new Put(ROWS[row]);
+ put.addColumn(essentialFamily, QUALIFIERS[0], row < numMatchingRows ?
VALUE : otherValue);
+ put.addColumn(joinedFamily, QUALIFIERS[0], VALUE);
+ puts.add(put);
+ }
+ table.put(puts);
+
+ SingleColumnValueFilter filter =
+ new SingleColumnValueFilter(essentialFamily, QUALIFIERS[0],
CompareOperator.EQUAL, VALUE);
+ // Makes the joined family non essential (see
SingleColumnValueFilter#isFamilyEssential),
+ // so that it is lazily populated through the joined heap for rows
passing the filter.
+ filter.setFilterIfMissing(true);
+ Scan scan = new Scan();
+ scan.setScanMetricsEnabled(true);
+ scan.setLoadColumnFamiliesOnDemand(true);
+ scan.setFilter(filter);
+
+ ResultScanner scanner = table.getScanner(scan);
+ int rowsReturned = 0;
+ for (Result result = scanner.next(); result != null; result =
scanner.next()) {
+ // Both the essential and the lazily loaded family must be present in
the result.
+ assertEquals(2, result.rawCells().length);
+ rowsReturned++;
+ }
+ scanner.close();
+ assertEquals(numMatchingRows, rowsReturned);
+ ScanMetrics metrics = scanner.getScanMetrics();
+ assertEquals(NUM_ROWS,
+
metrics.getCounter(ServerSideScanMetrics.COUNT_OF_ROWS_SCANNED_KEY_METRIC_NAME).get());
+ assertEquals(NUM_ROWS - numMatchingRows,
+
metrics.getCounter(ServerSideScanMetrics.COUNT_OF_ROWS_FILTERED_KEY_METRIC_NAME).get());
+ } finally {
+ TEST_UTIL.deleteTable(tableName);
+ }
+ }
+
@Test
public void testRowsFilteredMetric() throws Exception {
// Base scan configuration
