spark-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Jain, Abhishek 3. (Nokia - IN/Bangalore)" <abhishek.3.j...@nokia.com>
Subject Spark Metrics : Job Remains In "Running" State
Date Mon, 18 Mar 2019 13:15:07 GMT
Hi Team,

We are executing spark submit job by enabling the metrics(spark 2.4 on kubernetes) on the
user defined port(say 45010). We have observed that the job is not going into "Completed"
state even after it's completion(<sparkContext>.stop()).
The pods for this spark submit job remain in "Running" state.  I am able to collect the metrics
for both driver and executor/s on the defined port by using curl.

Below is the content of metrics.properties:
executor.sink.csv.class=org.apache.spark.metrics.sink.CsvSink
executor.sink.csv.period=1
executor.sink.csv.directory=/tmp/
executor.sink.csv.unit=seconds
driver.sink.csv.class=org.apache.spark.metrics.sink.CsvSink
driver.sink.csv.directory=/tmp/

*.sink.jmx.class=org.apache.spark.metrics.sink.JmxSink
driver.sink.jmx.period=1
driver.sink.jmx.unit=seconds

# Enable JVM metrics source for all instances by class name
driver.source.jvm.class=org.apache.spark.metrics.source.JvmSource
executor.source.jvm.class=org.apache.spark.metrics.source.JvmSource


Spark Submit Job:
export HADOOP_CONF_DIR=<xml files path>;sudo -E ./spark-submit --verbose --deploy-mode
cluster --master <k8s master host> --conf spark.app.name=<app Id> --conf spark.executor.instances=2
<program file>

Please let me know if it is the expected behavior ?

Regards,
Abhishek Jain

Mime
View raw message