Supervisor

Supervisor

Contents

1. Download and Configure Supervisor

1.1. Running Zookeeper with Supervisor

Supervisor monitors programs and automatically restarts them if they fail, which is useful for long-running tools like Flask, Zookeeper, and Storm.

## Download and Configure Supervisor

node:~$ sudo apt-get install supervisor

Next restart the supervisor service

node:~$ sudo service supervisor restart

Using Supervisor

Supervisor will now run any programs that have scripts in the /etc/supervisor/conf.d directory (see Zookeeper example in the next section).

Now we can run Supervisor with the various supervisorctl commands

node:~$ sudo supervisorctl help

A few useful commands are status, start, restart, and stop (you may need sudo).

For example, you can stop a specific process (e.g. Zookeeper) with:

node:~$ sudo supervisorctl stop zookeeper

This is the proper way to stop a process under supervision. If you start it again and check the status with

node:~$ sudo supervisorctl start zookeeper
node:~$ sudo supervisorctl status

you can see the process ids and uptime. If you try to kill the process without using supervisorctl, it will restart itself automatically (which is the whole point of Supervisor!)

You may find it useful to see the tail of a given log with:

node:~$ sudo supervisorctl tail zookeeper

Supervisor is useful for many applications like Kafka, Storm, or even running your Flask microserver. To add new programs simply and another configuration file (e.g. kafka.conf) to the conf.d directory and repeat the reread and update commands from above.

If you have the WebUI enabled, you can view and control Supervisor more conveniently by going to http://<node’s-public-dns>:9002. In particular, the “Clear Log” and “Tail -f” features are really nice for debugging.

## Running Zookeeper with Supervisor

First, if Zookeeper is still running from the Zookeeper Dev, stop it with:

node:~$ sudo /usr/local/zookeeper/bin/zkServer.sh stop

Write a configuration script in the /etc/supervisor/conf.d directory:

node:~$ sudo nano /etc/supervisor/conf.d/zookeeper.conf

With the following configuration

[program:zookeeper]
command=/usr/local/zookeeper/bin/zkServer.sh start-foreground
autorestart=true
user=root
stderr_logfile=/usr/local/zookeeper/zk.err.log
stdout_logfile=/usr/local/zookeeper/zk.out.log
stopasgroup=true
killasgroup=true
stopsignal=KILL

Now every time this node is started, the command following command will be ran, and importantly it will run this again if Zookeeper fails, rather than you checking it all the time and manually restarting. It also redirects standard error and output to the corresponding logs for easy troubleshooting.

In order to update Supervisor to read this new configuration file, you need to have it re-read the file and update it with the supervisorctl command

node:~$ sudo supervisorctl reread
node:~$ sudo supervisorctl update

That’s it, Supervisor will now keep Zookeeper up, even if unforeseen errors cause it to fail.