HBASE-26731 Add metrics for active and expired scanners (#4145)

Signed-off-by: Andrew Purtell <apurtell@apache.org>

Conflicts:
	hbase-server/src/main/java/org/apache/hadoop/hbase/regionserver/MetricsRegionServer.java
	hbase-server/src/test/java/org/apache/hadoop/hbase/regionserver/TestMetricsRegionServer.java
This commit is contained in:
Bryan Beaudreault 2022-03-04 17:03:32 -05:00 committed by Andrew Purtell
parent eb1c57d12e
commit fc92a00bd1
8 changed files with 47 additions and 2 deletions

View File

@ -226,6 +226,8 @@ public interface MetricsRegionServerSource extends BaseSource, JvmPauseMonitorSo
*/
void updateCompactionOutputSize(boolean isMajor, long bytes);
void incrScannerLeaseExpired();
// Strings used for exporting to metrics system.
String REGION_COUNT = "regionCount";
String REGION_COUNT_DESC = "Number of regions";
@ -592,4 +594,10 @@ public interface MetricsRegionServerSource extends BaseSource, JvmPauseMonitorSo
String BYTE_BUFF_ALLOCATOR_TOTAL_BUFFER_COUNT_DESC = "Total buffer count in ByteBuffAllocator";
String BYTE_BUFF_ALLOCATOR_USED_BUFFER_COUNT = "ByteBuffAllocatorUsedBufferCount";
String BYTE_BUFF_ALLOCATOR_USED_BUFFER_COUNT_DESC = "Used buffer count in ByteBuffAllocator";
String ACTIVE_SCANNERS = "activeScanners";
String ACTIVE_SCANNERS_DESC = "Gauge of currently active scanners";
String SCANNER_LEASE_EXPIRED_COUNT = "scannerLeaseExpiredCount";
String SCANNER_LEASE_EXPIRED_COUNT_DESC = "Count of scanners which were expired due to scanner lease timeout";
}

View File

@ -559,4 +559,6 @@ public interface MetricsRegionServerWrapper {
long getByteBuffAllocatorTotalBufferCount();
long getByteBuffAllocatorUsedBufferCount();
int getActiveScanners();
}

View File

@ -91,6 +91,8 @@ public class MetricsRegionServerSourceImpl
private final MetricHistogram pausesWithGc;
private final MetricHistogram pausesWithoutGc;
private final MutableFastCounter scannerLeaseExpiredCount;
public MetricsRegionServerSourceImpl(MetricsRegionServerWrapper rsWrap) {
this(METRICS_NAME, METRICS_DESCRIPTION, METRICS_CONTEXT, METRICS_JMX_CONTEXT, rsWrap);
}
@ -179,6 +181,8 @@ public class MetricsRegionServerSourceImpl
WARN_THRESHOLD_COUNT_DESC, 0L);
pausesWithGc = getMetricsRegistry().newTimeHistogram(PAUSE_TIME_WITH_GC_KEY);
pausesWithoutGc = getMetricsRegistry().newTimeHistogram(PAUSE_TIME_WITHOUT_GC_KEY);
scannerLeaseExpiredCount = getMetricsRegistry().newCounter(SCANNER_LEASE_EXPIRED_COUNT, SCANNER_LEASE_EXPIRED_COUNT_DESC, 0L);
}
@Override
@ -322,6 +326,11 @@ public class MetricsRegionServerSourceImpl
}
}
@Override
public void incrScannerLeaseExpired() {
scannerLeaseExpiredCount.incr();
}
/**
* Yes this is a get function that doesn't return anything. Thanks Hadoop for breaking all
* expectations of java programmers. Instead of returning anything Hadoop metrics expects
@ -582,7 +591,9 @@ public class MetricsRegionServerSourceImpl
rsWrap.getByteBuffAllocatorTotalBufferCount())
.addGauge(Interns.info(BYTE_BUFF_ALLOCATOR_USED_BUFFER_COUNT,
BYTE_BUFF_ALLOCATOR_USED_BUFFER_COUNT_DESC),
rsWrap.getByteBuffAllocatorUsedBufferCount());
rsWrap.getByteBuffAllocatorUsedBufferCount())
.addGauge(Interns.info(ACTIVE_SCANNERS, ACTIVE_SCANNERS_DESC),
rsWrap.getActiveScanners());
}
@Override

View File

@ -317,4 +317,8 @@ public class MetricsRegionServer {
public void incrementRegionSizeReportingChoreTime(long time) {
quotaSource.incrementRegionSizeReportingChoreTime(time);
}
public void incrScannerLeaseExpired() {
serverSource.incrScannerLeaseExpired();
}
}

View File

@ -661,6 +661,11 @@ class MetricsRegionServerWrapperImpl
return mobFileCacheHitRatio * 100;
}
@Override
public int getActiveScanners() {
return regionServer.getRpcServices().getScannersCount();
}
/**
* This is the runnable that will be executed on the executor every PERIOD number of seconds
* It will take metrics/numbers from all of the regions and use them to compute point in

View File

@ -534,6 +534,7 @@ public class RSRpcServices implements HBaseRPCErrorHandler, AdminService.Blockin
return;
}
LOG.info("Scanner lease {} expired {}", this.scannerName, rsh);
server.getMetrics().incrScannerLeaseExpired();
RegionScanner s = rsh.s;
HRegion region = null;
try {

View File

@ -138,6 +138,11 @@ public class MetricsRegionServerWrapperStub implements MetricsRegionServerWrappe
return 0;
}
@Override
public int getActiveScanners() {
return 0;
}
@Override
public long getReadRequestsCount() {
return 997;

View File

@ -251,7 +251,7 @@ public class TestMetricsRegionServer {
}
@Test
public void testServerQueryMeterSwitch() {
public void testTableQueryMeterSwitch() {
TableName tn1 = TableName.valueOf("table1");
// has been set disable in setUp()
rsm.updateReadQueryMeter(tn1, 500L);
@ -271,5 +271,14 @@ public class TestMetricsRegionServer {
rsm.updateWriteQueryMeter(tn1, 500L);
HELPER.assertGauge("ServerWriteQueryPerSecond_count", 500L, serverSource);
}
@Test
public void testScannerMetrics() {
HELPER.assertCounter("scannerLeaseExpiredCount", 0, serverSource);
rsm.incrScannerLeaseExpired();
HELPER.assertCounter("scannerLeaseExpiredCount", 1, serverSource);
HELPER.assertGauge("activeScanners", 0, serverSource);
}
}