Disk IO statistics are disabled by default. To enable disk IO statistics on a Cloudera Private Cloud Base cluster, you must set the file IO sampling
percentage to a non-zero value in the hdfs-site.xml
file.
-
Set the
dfs.datanode.fileio.profiling.sampling.percentage
property to a non-zero value in hdfs-site.xml
.
<property>
<name>dfs.datanode.fileio.profiling.sampling.percentage</name>
<value>25</value>
</property>
-
Access the disk IO statistics from the NameNode JMX page at
http://<namenode_host>:50070/jmx
.
In the following JMX output example, the time unit is milliseconds, and the
disk is healthy because the IO latencies are
low:
"name" : "Hadoop:service=DataNode,name=DataNodeVolume-/data/disk2/dfs/data/",
"modelerType" : "DataNodeVolume-/data/disk2/dfs/data/",
"tag.Context" : "dfs",
"tag.Hostname" : "n001.hdfs.example.com",
"TotalMetadataOperations" : 67,
"MetadataOperationRateAvgTime" : 0.08955223880597014,
...
"WriteIoRateNumOps" : 7321,
"WriteIoRateAvgTime" : 0.050812730501297636