Apache
/
hadoop
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

HDFS-15808. Add metrics for FSNamesystem read/write lock hold long time. (#2668) Contributed by tomscut. 

(cherry picked from commit 9cb51bf106)
This commit is contained in:
tomscut 2021-03-01 16:35:12 -08:00 committed by Konstantin V Shvachko
parent 1a33878279
commit c498ef4b6d
2 changed files with 48 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

14
hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/namenode/FSNamesystem.java
View File

@ -4758,6 +4758,20 @@ public class FSNamesystem implements Namesystem, FSNamesystemMBean,
return fsLock.getQueueLength();
}
@Metric(value = {"ReadLockLongHoldCount", "The number of time " +
"the read lock has been held for longer than the threshold"},
type = Metric.Type.COUNTER)
public long getNumOfReadLockLongHold() {
return fsLock.getNumOfReadLockLongHold();
}
@Metric(value = {"WriteLockLongHoldCount", "The number of time " +
"the write lock has been held for longer than the threshold"},
type = Metric.Type.COUNTER)
public long getNumOfWriteLockLongHold() {
return fsLock.getNumOfWriteLockLongHold();
}
int getNumberOfDatanodes(DatanodeReportType type) {
readLock();
try {

34
hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/namenode/FSNamesystemLock.java
View File

@ -108,6 +108,16 @@ class FSNamesystemLock {
private final AtomicReference<LockHeldInfo> longestReadLockHeldInfo =
new AtomicReference<>(new LockHeldInfo(0, 0, null));
private LockHeldInfo longestWriteLockHeldInfo = new LockHeldInfo(0, 0, null);
/**
* The number of time the read lock
* has been held longer than the threshold.
*/
private final AtomicLong numReadLockLongHold = new AtomicLong(0);
/**
* The number of time the write lock
* has been held for longer than the threshold.
*/
private final AtomicLong numWriteLockLongHold = new AtomicLong(0);
@VisibleForTesting
static final String OP_NAME_OTHER = "OTHER";
@ -176,6 +186,7 @@ class FSNamesystemLock {
final long readLockIntervalMs =
TimeUnit.NANOSECONDS.toMillis(readLockIntervalNanos);
if (needReport && readLockIntervalMs >= this.readLockReportingThresholdMs) {
numReadLockLongHold.incrementAndGet();
LockHeldInfo localLockHeldInfo;
do {
localLockHeldInfo = longestReadLockHeldInfo.get();
@ -253,6 +264,7 @@ class FSNamesystemLock {
LogAction logAction = LogThrottlingHelper.DO_NOT_LOG;
if (needReport &&
writeLockIntervalMs >= this.writeLockReportingThresholdMs) {
numWriteLockLongHold.incrementAndGet();
if (longestWriteLockHeldInfo.getIntervalMs() < writeLockIntervalMs) {
longestWriteLockHeldInfo =
new LockHeldInfo(currentTimeMs, writeLockIntervalMs,
@ -302,6 +314,28 @@ class FSNamesystemLock {
return coarseLock.writeLock().newCondition();
}
/**
* Returns the number of time the read lock
* has been held longer than the threshold.
*
* @return long - Number of time the read lock
* has been held longer than the threshold
*/
public long getNumOfReadLockLongHold() {
return numReadLockLongHold.get();
}
/**
* Returns the number of time the write lock
* has been held longer than the threshold.
*
* @return long - Number of time the write lock
* has been held longer than the threshold.
*/
public long getNumOfWriteLockLongHold() {
return numWriteLockLongHold.get();
}
/**
* Returns the QueueLength of waiting threads.
*