This is an automated email from the ASF dual-hosted git repository.
adoroszlai pushed a commit to branch master
in repository https://gitbox.apache.org/repos/asf/ozone.git
The following commit(s) were added to refs/heads/master by this push:
new 189fbdbe4c HDDS-12463. Add perf metrics in SCM for allocateBlock and
openPipeline (#8111)
189fbdbe4c is described below
commit 189fbdbe4c6f170e1064166e63d9b048ad33f752
Author: tanvipenumudy <[email protected]>
AuthorDate: Tue Apr 8 21:07:01 2025 +0530
HDDS-12463. Add perf metrics in SCM for allocateBlock and openPipeline
(#8111)
---
.../container/placement/metrics/SCMPerformanceMetrics.java | 14 ++++++++++++++
.../hadoop/hdds/scm/pipeline/PipelineManagerImpl.java | 3 +++
.../hadoop/hdds/scm/pipeline/SCMPipelineMetrics.java | 8 ++++++++
.../hadoop/hdds/scm/server/SCMBlockProtocolServer.java | 5 +++++
4 files changed, 30 insertions(+)
diff --git
a/hadoop-hdds/server-scm/src/main/java/org/apache/hadoop/hdds/scm/container/placement/metrics/SCMPerformanceMetrics.java
b/hadoop-hdds/server-scm/src/main/java/org/apache/hadoop/hdds/scm/container/placement/metrics/SCMPerformanceMetrics.java
index cd7a94ec6e..b2d142dc3d 100644
---
a/hadoop-hdds/server-scm/src/main/java/org/apache/hadoop/hdds/scm/container/placement/metrics/SCMPerformanceMetrics.java
+++
b/hadoop-hdds/server-scm/src/main/java/org/apache/hadoop/hdds/scm/container/placement/metrics/SCMPerformanceMetrics.java
@@ -51,6 +51,10 @@ public final class SCMPerformanceMetrics implements
MetricsSource {
private MutableRate deleteKeyFailureLatencyNs;
@Metric(about = "Latency for deleteKey success in nanoseconds")
private MutableRate deleteKeySuccessLatencyNs;
+ @Metric(about = "Latency for a successful allocateBlock call in nanoseconds")
+ private MutableRate allocateBlockSuccessLatencyNs;
+ @Metric(about = "Latency for a failed allocateBlock call in nanoseconds")
+ private MutableRate allocateBlockFailureLatencyNs;
public SCMPerformanceMetrics() {
this.registry = new MetricsRegistry(SOURCE_NAME);
@@ -78,6 +82,16 @@ public void getMetrics(MetricsCollector collector, boolean
all) {
deleteKeySuccessLatencyNs.snapshot(recordBuilder, true);
deleteKeyFailure.snapshot(recordBuilder, true);
deleteKeyFailureLatencyNs.snapshot(recordBuilder, true);
+ allocateBlockSuccessLatencyNs.snapshot(recordBuilder, true);
+ allocateBlockFailureLatencyNs.snapshot(recordBuilder, true);
+ }
+
+ public void updateAllocateBlockSuccessLatencyNs(long startNanos) {
+ allocateBlockSuccessLatencyNs.add(Time.monotonicNowNanos() - startNanos);
+ }
+
+ public void updateAllocateBlockFailureLatencyNs(long startNanos) {
+ allocateBlockFailureLatencyNs.add(Time.monotonicNowNanos() - startNanos);
}
public void updateDeleteKeySuccessStats(long startNanos) {
diff --git
a/hadoop-hdds/server-scm/src/main/java/org/apache/hadoop/hdds/scm/pipeline/PipelineManagerImpl.java
b/hadoop-hdds/server-scm/src/main/java/org/apache/hadoop/hdds/scm/pipeline/PipelineManagerImpl.java
index b5681aca96..fc8bb5432b 100644
---
a/hadoop-hdds/server-scm/src/main/java/org/apache/hadoop/hdds/scm/pipeline/PipelineManagerImpl.java
+++
b/hadoop-hdds/server-scm/src/main/java/org/apache/hadoop/hdds/scm/pipeline/PipelineManagerImpl.java
@@ -60,6 +60,7 @@
import org.apache.hadoop.metrics2.util.MBeans;
import org.apache.hadoop.ozone.ClientVersion;
import
org.apache.hadoop.ozone.common.statemachine.InvalidStateTransitionException;
+import org.apache.hadoop.util.Time;
import org.apache.ratis.protocol.exceptions.NotLeaderException;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
@@ -416,6 +417,7 @@ public int getNumberOfContainers(PipelineID pipelineID)
throws IOException {
@Override
public void openPipeline(PipelineID pipelineId)
throws IOException {
+ long startNanos = Time.monotonicNowNanos();
HddsProtos.PipelineID pipelineIdProtobuf = pipelineId.getProtobuf();
acquireWriteLock();
final Pipeline pipeline;
@@ -431,6 +433,7 @@ public void openPipeline(PipelineID pipelineId)
} finally {
releaseWriteLock();
}
+ metrics.updatePipelineCreationLatencyNs(startNanos);
metrics.incNumPipelineCreated();
metrics.createPerPipelineMetrics(pipeline);
}
diff --git
a/hadoop-hdds/server-scm/src/main/java/org/apache/hadoop/hdds/scm/pipeline/SCMPipelineMetrics.java
b/hadoop-hdds/server-scm/src/main/java/org/apache/hadoop/hdds/scm/pipeline/SCMPipelineMetrics.java
index 817bad4469..65d781943b 100644
---
a/hadoop-hdds/server-scm/src/main/java/org/apache/hadoop/hdds/scm/pipeline/SCMPipelineMetrics.java
+++
b/hadoop-hdds/server-scm/src/main/java/org/apache/hadoop/hdds/scm/pipeline/SCMPipelineMetrics.java
@@ -31,7 +31,9 @@
import org.apache.hadoop.metrics2.lib.Interns;
import org.apache.hadoop.metrics2.lib.MetricsRegistry;
import org.apache.hadoop.metrics2.lib.MutableCounterLong;
+import org.apache.hadoop.metrics2.lib.MutableRate;
import org.apache.hadoop.ozone.OzoneConsts;
+import org.apache.hadoop.util.Time;
/**
* This class maintains Pipeline related metrics.
@@ -54,6 +56,7 @@ public final class SCMPipelineMetrics implements
MetricsSource {
private @Metric MutableCounterLong numPipelineReportProcessed;
private @Metric MutableCounterLong numPipelineReportProcessingFailed;
private @Metric MutableCounterLong numPipelineContainSameDatanodes;
+ private @Metric MutableRate pipelineCreationLatencyNs;
private final Map<PipelineID, MutableCounterLong> numBlocksAllocated;
private final Map<PipelineID, MutableCounterLong> numBytesWritten;
@@ -100,6 +103,7 @@ public void getMetrics(MetricsCollector collector, boolean
all) {
numPipelineReportProcessed.snapshot(recordBuilder, true);
numPipelineReportProcessingFailed.snapshot(recordBuilder, true);
numPipelineContainSameDatanodes.snapshot(recordBuilder, true);
+ pipelineCreationLatencyNs.snapshot(recordBuilder, true);
numBytesWritten
.forEach((pid, metric) -> metric.snapshot(recordBuilder, true));
numBlocksAllocated
@@ -208,4 +212,8 @@ void incNumPipelineReportProcessingFailed() {
void incNumPipelineContainSameDatanodes() {
numPipelineContainSameDatanodes.incr();
}
+
+ public void updatePipelineCreationLatencyNs(long startNanos) {
+ pipelineCreationLatencyNs.add(Time.monotonicNowNanos() - startNanos);
+ }
}
diff --git
a/hadoop-hdds/server-scm/src/main/java/org/apache/hadoop/hdds/scm/server/SCMBlockProtocolServer.java
b/hadoop-hdds/server-scm/src/main/java/org/apache/hadoop/hdds/scm/server/SCMBlockProtocolServer.java
index 2b6eeb4626..1627342c29 100644
---
a/hadoop-hdds/server-scm/src/main/java/org/apache/hadoop/hdds/scm/server/SCMBlockProtocolServer.java
+++
b/hadoop-hdds/server-scm/src/main/java/org/apache/hadoop/hdds/scm/server/SCMBlockProtocolServer.java
@@ -192,6 +192,7 @@ public List<AllocatedBlock> allocateBlock(
String owner, ExcludeList excludeList,
String clientMachine
) throws IOException {
+ long startNanos = Time.monotonicNowNanos();
Map<String, String> auditMap = Maps.newHashMap();
auditMap.put("size", String.valueOf(size));
auditMap.put("num", String.valueOf(num));
@@ -235,17 +236,21 @@ public List<AllocatedBlock> allocateBlock(
AUDIT.logWriteFailure(buildAuditMessageForFailure(
SCMAction.ALLOCATE_BLOCK, auditMap, null)
);
+ perfMetrics.updateAllocateBlockFailureLatencyNs(startNanos);
} else {
AUDIT.logWriteSuccess(buildAuditMessageForSuccess(
SCMAction.ALLOCATE_BLOCK, auditMap));
+ perfMetrics.updateAllocateBlockSuccessLatencyNs(startNanos);
}
return blocks;
} catch (TimeoutException ex) {
+ perfMetrics.updateAllocateBlockFailureLatencyNs(startNanos);
AUDIT.logWriteFailure(buildAuditMessageForFailure(
SCMAction.ALLOCATE_BLOCK, auditMap, ex));
throw new IOException(ex);
} catch (Exception ex) {
+ perfMetrics.updateAllocateBlockFailureLatencyNs(startNanos);
AUDIT.logWriteFailure(buildAuditMessageForFailure(
SCMAction.ALLOCATE_BLOCK, auditMap, ex));
throw ex;
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
