The role of performers on the Spark master machine
Executors run only on nodes where this node has at least one working daemon, i.e. no executor will run in a node that will not serve as a worker.
However, where to start Master and Workers are based on your decision, there are no restrictions that Master and Worker cannot find together on the same node.
To start a working daemon on the same machine with your master, you can either edit the file conf/slaves
to add the master-ip to it, use start -all.sh during startup, or start anytime you want on the master node, start-slave.sh
and put the main url to Spark--master spark://master-host:7077
Update (based on suggestion from Daniel Darabos):
When you link to a tab Application Detail UI
Executors
, you can also find a line <driver>
for Executor ID
yours, then the driver it stands for is the process in which your job is scheduled and monitored, at startup main
, which you sent to the spark cluster, slicing your transformations and actions into RDD into stages, scheduling stages like TaskSets and organizing Executors
to complete tasks.
This one <driver>
will run in the node you call spark-submit
in client mode
, or one of the worker nodes incluster mode
source to share