The neo4j script under the
bin/ directory of any standard Neo4j install is the primary means of shutting down a running Neo4j instance. That script accepts a
stop argument that will try to shutdown the running instance by sending a
SIGTERM signal that the JVM interprets as a shutdown command and runs the installed shutdown hooks, which Neo4j implements to ensure a clean shutdown.
bin/neo4j stop a message appears saying that the process is stopping. Then it waits for the process to exit, by default for
2 minutes, controlled by the environmental variable
$NEO4J_SHUTDOWN_TIMEOUT (with the timeout set in seconds). The shell script will exit with a status code of 0 if the process exits before the
$NEO4J_SHUTDOWN_TIMEOUT period or it will print a failure message with the PID of the Neo4j process and will return status code
It is important to note that in either case, the Neo4j process will continue running. The reason is usually a checkpoint operation that takes longer than usual. If this happens, then the current state of the Neo4j process can still be inquired through use of the
bin/neo4j status command. This will return status code
0 if the process is running and print its
PID or it will return
3 if it’s not running.
The above allow two methods of checking the state of a Neo4j process that has been asked to shutdown.
One is to set the
$NEO4J_SHUTDOWN_TIMEOUT environment variable to something larger than
120 seconds. This is preferred if the expected time to shutdown is roughly known and the simplicity of a terminating command is desired.
The other is to use the exit code of the
bin/neo4j status command to execute a control loop after the
bin/neo4j stop command returns with a non
0 status (indicating a still running process). In this case, it may make sense to set the
$NEO4J_SHUTDOWN_TIMEOUT to something small, to jump to the control loop as soon as possible. For example