creep

Creep README file

Overview

Creep is an incremental deploy tool. It allows delta update from a local source (directory, Git repository or archive) to a FTP or SSH remote server. Its purpose is to deploy applications by incrementally copying local files to remote servers (e.g. production) with optional pre-processing.

Incremental deployment means Creep keeps track of deployed files on all remote locations. Only modified files are transferred between two deployments. This tracking mechanism depends on the type of directory used, for exemple Creep uses revision hashes when deploying from a Git repository.

Installing

You will need an environment with Python 3 (>= 3.6) before installing Creep.

Creep can be installed using either PIP or manual install from sources. If you choose to use PIP, just type the following:

$ pip3 install creep
$ creep -h # display help to ensure install worked properly

If you prefer manual install checkout the Git repository anywhere you want. Then create a symbolic link in your $PATH to file creep/creep.py. Last step is convenient but not mandatory, you can call Creep using full path to creep.py:

$ git clone https://github.com/r3c/creep.git
$ cd creep
$ sudo ln -s creep/creep.py /usr/bin/creep
$ creep -h # display help

After installation go to your project folder and continue reading to deploy your first project.

Quick start

First go to the directory you want to deploy, e.g. the dist folder of some project. It can be inside a Git repository or just be a regular folder, Creep will select a suitable default configuration either way. Create a new .creep.env file inside this directory with following JSON content:

$ cd path/to/your/dist/directory
$ cat > .creep.env << EOF
{
	"default": {
		"connection": "file:////tmp/creep-quickstart"
	}
}
EOF

Mind the quadruple slash in file:////tmp/creep-quickstart value, we'll see later why this is required. Once file is saved create the directory and execute Creep with no parameter:

$ mkdir /tmp/creep-quickstart
$ creep

Creep will notice you're deploying this project for the first time and ask you to confirm. Answer y to continue. It will display the full list of files in your project (by scanning file system or Git history) then ask you again to confirm. Enter y and Creep will deploy your project to directory /tmp/creep-quickstart.

Now if you try to execute Creep again you'll see a message saying no action is required. This is because deployment location now contains an up-to-date version of your project and Creep saved this information. Try to change some files (and git commit them if you were using Git) then execute the command again. This time Creep will copy only the file you changed rather than the full project.

This basic example shows how to incrementally deploy a project to some local directory. Next sections will show how to deploy to remote locations (FTP or SSH) and register several deployment configurations.

Environment file

As we saw during quick start steps, Creep reads deployment location(s) from a file called an environment file. It contains one or more named location(s) pointing to destinations you want to deploy to. By default Creep will search for an environment file named .creep.env in current directory, but we'll see later how this can be customized.

Environment configuration file uses JSON format and looks like this:

{
	"default": {
		"connection": "file:///../webdev/my-project"
	},
	"integration": {
		"connection": "ftp://me:password@my-dev-server/www-data"
	},
	"production": {
		"connection": "ssh://[e-mail]/www-data"
	}
}

Elements in the root object specify an available deployment location. Each one must have at least a connection string specifying protocol, address, credentials and/or path. Read details below for more information about supported protocols.

Once environment configuration file is ready you can start using Creep. Just type creep <name> where <name> is name of a configured location to initiate a deployment to this location. You can also specify multiple locations by running creep <name1> <name2> ... or use special location * to deploy everywhere (creep '*', don't forget to escape the * if you're running Creep from within a shell). If you don't specify a name Creep will deploy to location name default.

When invoked Creep will fetch last deployed revision from remote location and compute difference. When you deploy for the first time there is no last deployed revision so Creep will perform a full deploy. After each successful deployment it will save revision to remote location in a .creep.rev file.

Storing revision information on remote location keeps related data altogether and works well if you're not the only developer doing deployments. In case you prefer storing them locally, just add a new local option with value true for all affected locations in your .creep.env file:

{
	...
	"integration": {
		"connection": "ftp://me:password@my-dev-server/www-data/",
		"local": true
	},
	...
}

You can specify some options depending on the protocol you're using. To specify options just add a options JSON object property holding required options:

{
	...
	"default": {
		"connection": "ssh://www-data@localhost/www.mywebsite.com/",
		"options": {
			"extra": "-o StrictHostKeyChecking=no"
		}
	},
	...
}

Here is the list of supported protocols with expected connection string format and available options:

• Local file system:
  • Use connection format file:///path where path is relative to source directory.
  • Note the use of triple slash /// because file protocol has no hostname.
• FTP or FTPs:
  • Use connection format ftp://user:pass@host:port/path where user and pass are optional credentials (anonymous login will be used if they're missing), port is an optional FTP port which defaults to 21, and path is a path relative to FTP user home directory.
  • Use scheme ftps:// instead of ftp:// to enable TLS support.
  • Boolean option passive enables (default) or disables passive mode.
• SSH:
  • Use connection format ssh://user@host:port/path with same variables than the ones used for FTP deployment. No password can be specified here so you'll need to either enter password manually or setup SSH keys and start SSH agent.
  • String option extra can be used to pass parameters to SSH command as shown in example above.

Path is relative by default in all protocols. Add an extra slash / before your path to specify an absolute path, e.g. file:////opt/myproject or ssh://user@host//opt/myproject.

Note that environment files describe information about external locations and may contain passwords. For those reasons they should be excluded from your versionning and kept only on machines performing deployments.

It's often convenient to keep the environment file at the top-level of the directory you want to deploy, so that running creep without any argument will perform a deployment of this directory using settings from the .creep.env file it found there. You can create multiple environment files in sub-directories inside your projet in case they require different deployment configurations, for example one in src directory to deploy executable code to your application server and another one in assets directory to deploy static files to your web server.

Definition file

Creep supports another configuration file, called definition file. It's used to define how to detect changes in files and what preprocessing operations should be applied on files upon transfer. Create a new .creep.def file and put it in the same directories your environment files are. As opposed to environment files this one is bound to your project and should be shared along with other project files.

Definition configuration file uses JSON format and looks like this:

{
	"tracker": "hash",
	"options": {
		"algorithm": "md5",
		"follow": false
	},
	"modifiers": [
		{
			"pattern": "\\.dist$",
			"filter": "false"
		}
	]
}

The tracker property specifies how Creep should analyze differences between work directory and target location. When this option is not specified Creep will auto-detect it based on current environment. The options allows you to tune behavior of the selected tracker by specifying custom parameters. Here is the list of supported trackers and associated options:

• Git versionning:
  • Specify git tracker so local git executable is used to get diff between two revisions. When using this mode Creep relies on Git history and only needs to remember which revision has been deployed. It also allows you to manually specify the revision you want to deploy through command line argument.
  • No options are available for this tracker type.
• File hash:
  • Specify hash tracker to have Creep computing a hash of each file to detect differences. This mode has a higher overhead than Git since it has to save a value for each file rather than one unique revision, but can work with any regular folder.
  • String option algorithm selects the hashing algorithm to be used among sha1, sha256, sha512 or md5 (default).
  • Boolean option follow specifies whether symbolic links should be followed or ignored (default).

The modifiers part defines actions to perform on files before they're sent to remote locations (e.g. rename, compile, minify, obfuscate, etc.). Each modifier must define a regular expression pattern property to select files it affects, and processing actions that will be applied on them.

Here is an example of modifiers section in a definition configuration. Remember backslashes must be escaped in JSON, hence the double \\ used in regular expression patterns:

{
	...
	"modifiers": [
		{
			"pattern": "\\.dist$",
			"filter": "false"
		},
		{
			"pattern": "\\.min\\.js$"
		},
		{
			"pattern": "\\.js$",
			"modify": "uglifyjs --compress --mangle -- '{}'"
		},
		{
			"pattern": "(.*)\\.less$",
			"rename": "\\1.css",
			"modify": "lessc --clean-css '{}'",
			"link": "find . -name '*.less' | xargs grep -Fl \"$(basename '{}')\" || true"
		}
	]
	...
}

Creep evaluates modifiers in sequence and one file can only match one modifier: evaluation stops after the first matched one. For each file matching a modifier, associated properties (if any) are applied to it. Available modifier properties are listed below by execution order:

• rename property specifies a new name for file and supports \\n back references on groups from the regular expression used in pattern (remember backslashes must be escaped in JSON).
  • In the example above, files ending with .less will have their extension changed to .css: the back reference \\1 captured original file name without extension in associated pattern.
• link property specifies a shell command expected to output path to all files that must also be included in the deployment along with matched file. Command can contain special {} token which will be replaced by absolute path to matched file. Output must contain one path per line and each path must be relative to source directory.
  • In the example above, a command using find and grep is used to list all files referencing currently ones, so they're also sent to remote location whenever the file they reference is changed.
  • Note the regular expression could have been more specific, but the point is to be sure to include all changed files when deploying ; a few false positives will just cause a harmless extra cost due to additional files being deployed while not changed.
• modify property specifies a shell command expected to output replacement contents for the file being sent to remote location. It supports {} token similar to the link property above.
  • In the example above, executable uglifyjs is called to minify JavaScript files (ending with .js) ; uglifyjs prints result to standard output, which is then used to overwrite contents of matched JavaScript files.
  • Note presence of a rule which matches files with name ending with .min.js: it doesn't specify any action but prevents the \.js$ rule from being triggered for files that are already minified.
• chmod property specifies the file permission to be applied in octal format. Mode "0644" (read/write for owner, read-only for everyone else) is used if this property is not set.
• filter property specifies a shell command used to decide whether file should be excluded from deployment or not. It supports {} token similar to the link property above. If command execution returns a non-zero exit code then file won't be sent to remote location.
  • In the example above, the false command is used to exclude files with name ending with .dist from deployment.
  • Empty string value can also be used to always exclude files. It's equivalent to the false command used in the example above but has better portability.

Creep always appends two modifiers to filter to exclude environment and definition files from deployments. You shouldn't need to change this behavior, but you may do so by adding explicit modifiers matching them.

You can also specify definition configuration as a JSON string instead of file using -d command line option:

creep -d '{"tracker": "hash"}'

Troubleshooting

This project is still under develpement and may not behave as you would expect. In case of issue the -v (verbose) switch may help you understanding how your environment and definition files are used.

If you can't figure out what's happening don't hesitate to open an issue on GitHub or contact me!