Super simple Node.js deployment tool
Core API JSCoverage: 95.52%
Pod simplifies the workflow of setting up, updating and managing multiple Node.js apps on a Linux server. Perfect for hosting personal Node stuff on a VPS. There are essentially two parts: 1.
git push deploy (by using git hooks) and 2. process management (by using pm2)
It doesn't manage DNS routing for you (personally I'm doing that in Nginx) but you can use pod to run a node-http-proxy server on port 80 that routes incoming requests to other apps.
On the server:
$ pod create myapp
On your local machine:
$ git clone ssh://your-server/pod_dir/myapp.git# hack hack hack, commit# make sure your main file is app.js# or specify "main" in package.json$ git push
You can also just add it as a remote to an existing local repo:
$ git remote add deploy ssh://your-server/pod_dir/myapp.git$ git push deploy master
That's it! App should be automatically running after the push. For later pushes, app process will be restarted. There's more to it though, read on to find out more.
$ [sudo] npm install -g pod
To make pod auto start all managed apps on system startup, you might also want to write a simple upstart script that contains something like this:
# /etc/init/pod.confstart on (local-filesystems and net-device-up IFACE!=lo)exec sudo -u <username> /path/to/node /path/to/pod startall
The first time you run
pod it will ask you where you want to put your stuff. The structure of the given directory will look like this:
.├── repos # holds the bare .git repos│ └── example.git└── apps # holds the working copies└── example├──app.js└──.podhook
Usage: pod [command]Commands:create <app> Create a new appremote <app> <repo> Create a app from a remote GitHub reporm <app> Delete an appstart <app> Start an app monitored by pm2stop <app> Stop an apprestart <app> Restart an app that's already runninglist List apps and statusstartall Start all apps not already runningstopall Stop all appsrestartall Restart all running appsprune Clean up dead fileshooks Update hooks after a pod upgradeweb [command] Start/stop/restart the web interfacehelp You are reading it right now
$ pod web [stop|restart|status]
This command will start the pod web service, by default at port 19999, which provides several functionalities:
/: a web interface that displays current apps status.
/json: returns app status data in json format.
/jsonp: accepts jsonp. This route must be enabled in config.
/hooks/appname: trigger fetch/restart for corresponding remote apps.
/json require a basic http authentication. Make sure to set the username and password in the config file.
You can setup an app to track a remote GitHub repo by using the
pod remote command:
$ pod remote my-remote-app username/repo
After this, add a webhook to your GitHub repo pointing at your web interface's
/hooks/my-remote-app. The webhook will trigger a fetch and restart just like local apps. By default a remote app will be tracking the master branch only, if you want to track a different branch, you can change it in the config file.
You can also set up a remote app to track an arbitrary git address. However in that case you need to manually make a POST request conforming to the GitHub webhook payload.
Starting in 0.8.2, GitLab webhooks are also supported.
Starting in 0.8.6, Bitbucket webhooks are also supported.
The config file lives at
~/.podrc. Note since 0.7.0 all fields follow the underscore format so check your config file if things break after upgrading.
// where pod puts all the stuff"root": "/srv"// default env"node_env": "development"// this can be overwritten in each app's package.json's "main" field// or in the app's configuration below using the "script" field"default_script": "app.js"// minimum uptime to be considered stable,// in milliseconds. If not set, all restarts// are considered unstable."min_uptime": 3600000// max times of unstable restarts allowed// before the app is auto stopped."max_restarts": 10// config for the web interface"web":// set these! default is admin/admin"username": "admin""password": "admin""port": 19999// allow jsonp for web interface, defaults to false"jsonp": true"apps":"example1":// passed to the app as process.env.NODE_ENV// if not set, will inherit from global settings"node_env": "production"// passed to the app as process.env.PORT// if not set, pod will try to parse from app's// main file (for displaying only), but not// guarunteed to be correct."port": 8080// pod will look for this script before checking// in package.json of the app."script": "dist/server.js"// *** any valid pm2 config here gets passed to pm2. ***// spin up 2 instances using cluster module"instances": 2// pass in additional command line args to the app"args": "['--toto=heya coco', '-d', '1']"// file paths for stdout, stderr logs and pid.// will be in ~/.pm2/ if not specified"error_file": "/absolute/path/to/stderr.log""out_file": "/absolute/path/to/stdout.log""example2":// you can override global settings"min_uptime": 1000"max_restarts": 200"my-remote-app":"remote": "yyx990803/my-remote-app" // github shorthand"branch": "test" // if not specified, defaults to master// pass environment variables to all apps"env":"SERVER_NAME": "Commodore""CERT_DIR": "/path/to/certs"
Pod relies on pm2 for process management under the hood. When installing pod, the
pm2 executable will also be linked globally. You can invoke
pm2 commands for more detailed process information.
Logging is delegated to
pm2. If you didn't set an app's
error_file options, logs will default to be saved at
If things go wrong and restarting is not fixing them, try
pm2 kill. It terminates all pm2-managed processes and resets potential env variable problems.
All pod commands only concerns apps present in pod's config file, so it's fine if you use pm2 separately to run additional processes.
By default pod will run
npm install for you everytime you push to the repo. To override this behavior and run custom shell script before restarting the app, just include a
.podhook file in your app. If
.podhook exits with code other than 0, the app will not be restarted and will hard reset to the commit before the push.
component installnpm installgrunt buildgrunt testpassed=$?if [[ $passed != 0 ]]; then# test failed, exit. app's working tree on the server will be reset.exit $passedfi# restart is automatic so no need to include that here
You can also directly edit the post-receive script of an app found in
pod-root-dir/repos/my-app.git/hooks/post-receive if you wish.
NOTE: the API can only be used after you've initiated the config file via command line.
require('pod') will return the API. You have to wait till it's ready to do anything with it:
var pod =pod
The API methods follow a conventional error-first callback style. Refer to the source for more details.
Ready to go docker images:
.podrcby using the
readlinemodule blocking stdin. This caused issues when attempting to clone a repository that required a username/password.
strip()function so it processes github ssh urls correctly. (Thanks to @mathisonian)
package.jsonis now more npm compliant. (e.g. it now allows omitting file extensions)
default_script. Consult the configuration section for more details.
pod remotecommands. See web interface and using a remote github repo for more details.
git fetch --all+
git reset --hard origin/masterinstead of a pull. This allows users to do forced pushes that isn't necesarrily ahead of the working copy.
pod hookscommands. Make sure to run
pod hooksafter upgrading pod, as you will want to update the hooks that are already created in your existing apps.