A module to run and supervise background processes from Caddy
For every process in the supervisor caddyfile directive a command is executed in background and killed when caddy stops.
{
# Must be in global options
supervisor {
php-fpm --no-daemonize {
dir /path/to/desired/working-dir # default to current dir
env APP_ENV production
env DEBUG false
user www-data
restart_policy always # default to 'always', other values allowed: 'never', 'on_failure'
redirect_stdout file /var/log/fpm.log # redirect command stdout to a file. Default to caddy `stdout`
redirect_stderr file /var/log/fpm-error.log # redirect command stderr to a file. Default to caddy `stderr`
termination_grace_period 30s # default to '10s', amount of time to wait for application graceful termination before killing it
replicas 3 # default to 1, number of instances that should be executed
}
# block configuration is optional
node worker.js
}
}
mysite.com
- command: the command to be executed. Supports template.
- dir: the working directory the command should be executed in. Supports template.
- env: declare environment variable that should be passed to command. This property can be repeated. Supports template.
- redirect_stdout: redirect command stdout. Default:
stdout
, Possible values:- null: discard output
- stdout: redirect to the caddy process stdout
- stderr: redirect to the caddy process stderr
- file /path/to/file: redirect output to a file
- redirect_stderr: redirect command stderr. Default:
stderr
, See above for possible values. - restart_policy: define under which conditions the command should be restarted after exit. Default:
always
Valid values:- never: do not restart the command
- on_failure: restart if exit code is not 0
- always: always restart
- termination_grace_period: amount of time to wait for application graceful termination before killing it. Ex: 10s
- replicas: number of instances that should be executed. Default: 1.
- user: the user that runs the command. Ignored on windows systems
On windows termination_grace_period is ignored and the command is killed immediatelly due to lack of signals support.
To enable different configuration per replica, you can use go templates on the fields marked with Supports template".
The following information are available to templates:
- Replica: the index of the current replica, starting from 0
Templates also supports all functions from http://masterminds.github.io/sprig/
Example:
{
supervisor {
myapp --port "{{add 8000 .Replica}}" {
replicas 5
}
}
}
reverse_proxy * localhost:8000-8004
To avoid spending too many resources on a crashing application, this plugin makes use of exponential backoff.
That means that when the command fail, it will be restarted with a delay of 0 seconds. If it fails again it will be restarted with a delay of 1 seconds, then on every sucessive failure the delay time doubles, with a max limit of 5 minutes.
If the command runs stable for at least 10 minutes, the restart delay is reset to 0 seconds.
PHP server (useful if you want a single container PHP application with php-fpm):
{
supervisor {
php-fpm
}
}
example.com
php_fastcgi 127.0.0.1:9000
root * .
encode gzip
file_server
AspNet Core application on windows:
{
supervisor {
dotnet ./MyApplication.dll {
env ASPNETCORE_URLS http://localhost:5000
dir "C:\MyApplicationFolder"
redirect_stdout stdout
redirect_stderr stderr
restart_policy always
}
}
}
example.com
reverse_proxy localhost:5000
Use the xcaddy
tool to build a version of caddy with this module:
xcaddy build \
--with github.com/baldinof/caddy-supervisor
- Dedicated Caddyfile support (Supervisorfile)
- Send processes outputs to Caddy logs
This package is continuation of https://github.com/lucaslorentz/caddy-supervisor which only supports Caddy v1.
Thank you @lucaslorentz for the original work ❤️