Implementation:Apache Spark Stop All

Field	Value
Source	Repo Apache Spark
Domains	Deployment
Type	API Doc
Related	Principle:Apache_Spark_Cluster_Lifecycle_Shutdown

Overview

Shell scripts that stop all Spark standalone cluster daemons in the correct dependency order.

Description

sbin/stop-all.sh orchestrates full cluster shutdown by:

• Calling stop-workers.sh first to terminate all worker daemons
• Then calling stop-master.sh to terminate the master daemon
• Supporting the --wait flag to block until all workers have fully terminated

Individual scripts can be used independently for partial shutdowns:

• sbin/stop-master.sh -- stops only the master daemon
• sbin/stop-workers.sh -- stops only the worker daemons across all hosts
• sbin/decommission-worker.sh -- sends SIGPWR to allow graceful task completion before worker shutdown

Usage

Run on the master node to shut down the entire cluster. For rolling maintenance, use decommission-worker.sh on individual workers to allow in-flight tasks to complete before the worker exits.

Code Reference

Source: sbin/stop-all.sh (L1-49), sbin/stop-master.sh (L1-29), sbin/stop-workers.sh (L1-29), sbin/decommission-worker.sh (L1-58)

Signatures:

# Stop all daemons (workers first, then master)
sbin/stop-all.sh [--wait]

# Gracefully decommission a worker (drain tasks first)
sbin/decommission-worker.sh [--block-until-exit]

I/O

Direction	Description
Inputs	PID files from daemon startup, --wait flag (optional)
Outputs	All daemons stopped, PID files cleaned up

Examples

Stop all cluster daemons:

./sbin/stop-all.sh

Stop all and wait for full termination:

./sbin/stop-all.sh --wait

Gracefully decommission a single worker:

./sbin/decommission-worker.sh --block-until-exit

Stop only the master daemon:

./sbin/stop-master.sh

Related

• Principle:Apache_Spark_Cluster_Lifecycle_Shutdown