[
https://issues.apache.org/jira/browse/HDFS-14313?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16887592#comment-16887592
]
Yiqun Lin commented on HDFS-14313:
----------------------------------
I have two comments for the core implementation of the
ReplicaCachingGetSpaceUsed.
{code}
@Override
public synchronized Set<? extends Replica> deepCopyReplica(String bpid)
throws IOException {
Set<? extends Replica> replicas =
new HashSet<>(volumeMap.replicas(bpid) == null ? Collections.EMPTY_SET
: volumeMap.replicas(bpid));
return replicas;
}
{code}
why not use dataset lock? Here the volumeMap is thread-safe controlled by the
dataset lock in FsDatasetImpl operations. Synchronized cannot ensure the race
condition. Also it would be better to use {{ Collections.unmodifiableSet}} to
make replica info is not allowed to be modified outside.
{code}
if (CollectionUtils.isNotEmpty(replicaInfos)) {
+ for (ReplicaInfo replicaInfo : replicaInfos) {
+ if (Objects.equals(replicaInfo.getVolume().getStorageID(),
+ volume.getStorageID())) {
+ dfsUsed += replicaInfo.getBytesOnDisk();
+ dfsUsed += replicaInfo.getMetadataLength();
+ count++;
+ }
+ }
+ }
{code}
I think {{replicaInfo.getBlockDataLength}} will be more accurate than
{{replicaInfo.getBytesOnDisk}}. It gets the length from actual file while
getBytesOnDisk is from block inner variable.
Will take detail review soon.
> Get hdfs used space from FsDatasetImpl#volumeMap#ReplicaInfo in memory
> instead of df/du
> ----------------------------------------------------------------------------------------
>
> Key: HDFS-14313
> URL: https://issues.apache.org/jira/browse/HDFS-14313
> Project: Hadoop HDFS
> Issue Type: Improvement
> Components: datanode, performance
> Affects Versions: 2.6.0, 2.7.0, 2.8.0, 2.9.0, 3.0.0, 3.1.0
> Reporter: Lisheng Sun
> Assignee: Lisheng Sun
> Priority: Major
> Attachments: HDFS-14313.000.patch, HDFS-14313.001.patch,
> HDFS-14313.002.patch, HDFS-14313.003.patch, HDFS-14313.004.patch,
> HDFS-14313.005.patch, HDFS-14313.006.patch
>
>
> There are two ways of DU/DF getting used space that are insufficient.
> # Running DU across lots of disks is very expensive and running all of the
> processes at the same time creates a noticeable IO spike.
> # Running DF is inaccurate when the disk sharing by multiple datanode or
> other servers.
> Getting hdfs used space from FsDatasetImpl#volumeMap#ReplicaInfos in memory
> is very small and accurate.
--
This message was sent by Atlassian JIRA
(v7.6.14#76016)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
