[
https://issues.apache.org/jira/browse/HADOOP-19362?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17905220#comment-17905220
]
ASF GitHub Bot commented on HADOOP-19362:
-----------------------------------------
KeeProMise commented on code in PR #7224:
URL: https://github.com/apache/hadoop/pull/7224#discussion_r1882374273
##########
hadoop-common-project/hadoop-common/src/main/java/org/apache/hadoop/ipc/Server.java:
##########
@@ -668,9 +675,41 @@ void updateMetrics(Call call, long
processingStartTimeNanos, boolean connDropped
}
}
- void updateDeferredMetrics(String name, long processingTime) {
+ /**
+ * Update rpc metrics for defered calls.
+ * @param call The Rpc Call
+ * @param name Rpc method name
+ * @param processingTime processing call in ms unit.
+ */
+ void updateDeferredMetrics(Call call, String name, long processingTime) {
+ long completionTimeNanos = Time.monotonicNowNanos();
+ long arrivalTimeNanos = call.timestampNanos;
+
+ ProcessingDetails details = call.getProcessingDetails();
+ long waitTime =
+ details.get(Timing.LOCKWAIT, rpcMetrics.getMetricsTimeUnit());
+ long responseTime =
+ details.get(Timing.RESPONSE, rpcMetrics.getMetricsTimeUnit());
+ rpcMetrics.addRpcLockWaitTime(waitTime);
+ rpcMetrics.addRpcProcessingTime(processingTime);
+ rpcMetrics.addRpcResponseTime(responseTime);
rpcMetrics.addDeferredRpcProcessingTime(processingTime);
rpcDetailedMetrics.addDeferredProcessingTime(name, processingTime);
+ // don't include lock wait for detailed metrics.
+ processingTime -= waitTime;
+ rpcDetailedMetrics.addProcessingTime(name, processingTime);
+
+ // Overall processing time is from arrival to completion.
+ long overallProcessingTime = rpcMetrics.getMetricsTimeUnit()
+ .convert(completionTimeNanos - arrivalTimeNanos, TimeUnit.NANOSECONDS);
+ rpcDetailedMetrics.addOverallProcessingTime(name, overallProcessingTime);
+ callQueue.addResponseTime(name, call, details);
+ if (isLogSlowRPC()) {
+ logSlowRpcCalls(name, call, details);
+ }
+ if (details.getReturnStatus() == RpcStatusProto.SUCCESS) {
+ rpcMetrics.incrRpcCallSuccesses();
+ }
Review Comment:
There is some duplicate code in this method and updateMetrics; we can
extract the common code.
##########
hadoop-common-project/hadoop-common/src/main/java/org/apache/hadoop/ipc/ProtobufRpcEngine2.java:
##########
@@ -425,28 +426,39 @@ static class ProtobufRpcEngineCallbackImpl
private final RPC.Server server;
private final Call call;
private final String methodName;
- private final long setupTime;
+ private final long callStartNanos;
Review Comment:
Can callStartNanos be made an attribute of the call object, so that thread
variables are not needed?
> RPC metrics should be updated correctly when call is defered.
> -------------------------------------------------------------
>
> Key: HADOOP-19362
> URL: https://issues.apache.org/jira/browse/HADOOP-19362
> Project: Hadoop Common
> Issue Type: Improvement
> Components: common
> Reporter: farmmamba
> Assignee: farmmamba
> Priority: Major
> Labels: pull-request-available
>
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
