[
https://issues.apache.org/jira/browse/HDFS-16971?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17711137#comment-17711137
]
ASF GitHub Bot commented on HDFS-16971:
---------------------------------------
goiri commented on code in PR #5536:
URL: https://github.com/apache/hadoop/pull/5536#discussion_r1163407205
##########
hadoop-common-project/hadoop-common/src/main/java/org/apache/hadoop/fs/FileSystem.java:
##########
@@ -4252,6 +4259,14 @@ public void incrementBytesReadByDistance(int distance,
long newBytes) {
}
}
+ /**
+ * Increment the time taken to read bytes from remote in the statistics.
+ * @param duration time taken in ms to read bytes from remote
+ */
+ public void increaseRemoteBytesReadTime(long duration) {
Review Comment:
`final long durationMs`
##########
hadoop-hdfs-project/hadoop-hdfs-client/src/main/java/org/apache/hadoop/hdfs/DFSClient.java:
##########
@@ -3090,10 +3090,14 @@ public Peer newConnectedPeer(InetSocketAddress addr,
}
}
- void updateFileSystemReadStats(int distance, int nRead) {
+ void updateFileSystemReadStats(int distance, int nRead, long readTimeMS) {
Review Comment:
As we are at it, `readBytes`?
##########
hadoop-hdfs-project/hadoop-hdfs-client/src/main/java/org/apache/hadoop/hdfs/StripeReader.java:
##########
@@ -351,9 +351,12 @@ void readStripe() throws IOException {
// first read failure
while (!futures.isEmpty()) {
try {
+ long beginRead = Time.monotonicNow();
Review Comment:
`beginReadMs`?
##########
hadoop-common-project/hadoop-common/src/main/java/org/apache/hadoop/fs/FileSystem.java:
##########
@@ -3942,6 +3942,7 @@ public static class StatisticsData {
private volatile long bytesReadDistanceOfThreeOrFour;
private volatile long bytesReadDistanceOfFiveOrLarger;
private volatile long bytesReadErasureCoded;
+ private volatile long remoteBytesReadTimeMS;
Review Comment:
I'm not sure "Bytes" makes sense in the name.
`remoteReadTimeMS`?
##########
hadoop-common-project/hadoop-common/src/test/java/org/apache/hadoop/fs/TestFileSystemStorageStatistics.java:
##########
@@ -74,6 +75,7 @@ public void setup() {
statistics.incrementBytesReadByDistance(1, RandomUtils.nextInt(0, 100));
statistics.incrementBytesReadByDistance(3, RandomUtils.nextInt(0, 100));
statistics.incrementBytesReadErasureCoded(RandomUtils.nextInt(0, 100));
+ statistics.increaseRemoteBytesReadTime(RandomUtils.nextInt(0, 100));
Review Comment:
Can we assert something?
> Add read time metrics for remote reads in Statistics
> ----------------------------------------------------
>
> Key: HDFS-16971
> URL: https://issues.apache.org/jira/browse/HDFS-16971
> Project: Hadoop HDFS
> Issue Type: Improvement
> Components: hdfs
> Reporter: Melissa You
> Priority: Minor
> Labels: pull-request-available
> Fix For: 3.3.0, 3.4.0
>
>
> In Filesystem.java, currently it only collects bytes read not the
> corresponding timing of reads.
> In particular, we're interested in time spent on remote reads(not local
> reads). Adding timing info will help us understand it better and further
> analyze how locality of DN behaves.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
