Files
cloudron-box/src/docker.js
girish@cloudron.io b9d4b8f6e8 Remove docker images by tag
docker pull previously used to pull down all tags.
    docker pull tag1 # might pull down tag2, tag3 if they are all same!
    docker rm tag1 # oops, tag2 and tag3 still hold on to the image

However, the above feature was not possible with registry v2 (some
technical stuff to do with each tag being separately signed). As a result,
this feature was removed from v1 as well - https://github.com/docker/docker/pull/10571

This means we can now do:
    docker pull tag1 # nice
    docker rm tag1 # image goes away if no one else is using it

references:
https://github.com/docker/docker/issues/8689
https://github.com/docker/docker/pull/8193 (added this feature to v1)
https://github.com/docker/docker/issues/8141 (the request)
https://github.com/docker/docker/pull/10571 (removes the v1 feature)

Fixes #563
2016-01-21 15:53:51 -08:00

349 lines
13 KiB
JavaScript

'use strict';
var addons = require('./addons.js'),
async = require('async'),
assert = require('assert'),
config = require('./config.js'),
debug = require('debug')('box:src/docker.js'),
Docker = require('dockerode'),
safe = require('safetydance'),
semver = require('semver'),
util = require('util'),
_ = require('underscore');
exports = module.exports = {
connection: connectionInstance(),
downloadImage: downloadImage,
createContainer: createContainer,
startContainer: startContainer,
stopContainer: stopContainer,
stopContainerByName: stopContainer,
stopContainers: stopContainers,
deleteContainer: deleteContainer,
deleteContainerByName: deleteContainer,
deleteImage: deleteImage,
deleteContainers: deleteContainers,
createSubcontainer: createSubcontainer
};
function connectionInstance() {
var docker;
if (process.env.BOX_ENV === 'test') {
// test code runs a docker proxy on this port
docker = new Docker({ host: 'http://localhost', port: 5687 });
// proxy code uses this to route to the real docker
docker.options = { socketPath: '/var/run/docker.sock' };
} else {
docker = new Docker({ socketPath: '/var/run/docker.sock' });
}
return docker;
}
function debugApp(app, args) {
assert(!app || typeof app === 'object');
var prefix = app ? (app.location || '(bare)') : '(no app)';
debug(prefix + ' ' + util.format.apply(util, Array.prototype.slice.call(arguments, 1)));
}
function targetBoxVersion(manifest) {
if ('targetBoxVersion' in manifest) return manifest.targetBoxVersion;
if ('minBoxVersion' in manifest) return manifest.minBoxVersion;
return '99999.99999.99999'; // compatible with the latest version
}
function pullImage(manifest, callback) {
var docker = exports.connection;
docker.pull(manifest.dockerImage, function (err, stream) {
if (err) return callback(new Error('Error connecting to docker. statusCode: %s' + err.statusCode));
// https://github.com/dotcloud/docker/issues/1074 says each status message
// is emitted as a chunk
stream.on('data', function (chunk) {
var data = safe.JSON.parse(chunk) || { };
debug('pullImage %s: %j', manifest.id, data);
// The information here is useless because this is per layer as opposed to per image
if (data.status) {
} else if (data.error) {
debug('pullImage error %s: %s', manifest.id, data.errorDetail.message);
}
});
stream.on('end', function () {
debug('downloaded image %s of %s successfully', manifest.dockerImage, manifest.id);
var image = docker.getImage(manifest.dockerImage);
image.inspect(function (err, data) {
if (err) return callback(new Error('Error inspecting image:' + err.message));
if (!data || !data.Config) return callback(new Error('Missing Config in image:' + JSON.stringify(data, null, 4)));
if (!data.Config.Entrypoint && !data.Config.Cmd) return callback(new Error('Only images with entry point are allowed'));
debug('This image of %s exposes ports: %j', manifest.id, data.Config.ExposedPorts);
callback(null);
});
});
stream.on('error', function (error) {
debug('error pulling image %s of %s: %j', manifest.dockerImage, manifest.id, error);
callback(error);
});
});
}
function downloadImage(manifest, callback) {
assert.strictEqual(typeof manifest, 'object');
assert.strictEqual(typeof callback, 'function');
debug('downloadImage %s %s', manifest.id, manifest.dockerImage);
var attempt = 1;
async.retry({ times: 10, interval: 15000 }, function (retryCallback) {
debug('Downloading image %s %s. attempt: %s', manifest.id, manifest.dockerImage, attempt++);
pullImage(manifest, function (error) {
if (error) console.error(error);
retryCallback(error);
});
}, callback);
}
function createSubcontainer(app, name, cmd, options, callback) {
assert.strictEqual(typeof app, 'object');
assert.strictEqual(typeof name, 'string');
assert(!cmd || util.isArray(cmd));
assert.strictEqual(typeof options, 'object');
assert.strictEqual(typeof callback, 'function');
var docker = exports.connection,
isAppContainer = !cmd;
var manifest = app.manifest;
var developmentMode = !!manifest.developmentMode;
var exposedPorts = {}, dockerPortBindings = { };
var stdEnv = [
'CLOUDRON=1',
'WEBADMIN_ORIGIN=' + config.adminOrigin(),
'API_ORIGIN=' + config.adminOrigin(),
'APP_ORIGIN=https://' + config.appFqdn(app.location),
'APP_DOMAIN=' + config.appFqdn(app.location)
];
// docker portBindings requires ports to be exposed
exposedPorts[manifest.httpPort + '/tcp'] = {};
dockerPortBindings[manifest.httpPort + '/tcp'] = [ { HostIp: '127.0.0.1', HostPort: app.httpPort + '' } ];
var portEnv = [];
for (var e in app.portBindings) {
var hostPort = app.portBindings[e];
var containerPort = manifest.tcpPorts[e].containerPort || hostPort;
exposedPorts[containerPort + '/tcp'] = {};
portEnv.push(e + '=' + hostPort);
dockerPortBindings[containerPort + '/tcp'] = [ { HostIp: '0.0.0.0', HostPort: hostPort + '' } ];
}
var memoryLimit = manifest.memoryLimit || (developmentMode ? 0 : 1024 * 1024 * 200); // 200mb by default
// for subcontainers, this should ideally be false. but docker does not allow network sharing if the app container is not running
// this means cloudron exec does not work
var isolatedNetworkNs = true;
addons.getEnvironment(app, function (error, addonEnv) {
if (error) return callback(new Error('Error getting addon environment : ' + error));
var containerOptions = {
name: name, // used for filtering logs
// do _not_ set hostname to app fqdn. doing so sets up the dns name to look up the internal docker ip. this makes curl from within container fail
// for subcontainers, this should not be set because we already share the network namespace with app container
Hostname: isolatedNetworkNs ? (semver.gte(targetBoxVersion(app.manifest), '0.0.77') ? app.location : config.appFqdn(app.location)) : null,
Tty: isAppContainer,
Image: app.manifest.dockerImage,
Cmd: (isAppContainer && developmentMode) ? [ '/bin/bash', '-c', 'echo "Development mode. Use cloudron exec to debug. Sleeping" && sleep infinity' ] : cmd,
Env: stdEnv.concat(addonEnv).concat(portEnv),
ExposedPorts: isAppContainer ? exposedPorts : { },
Volumes: { // see also ReadonlyRootfs
'/tmp': {},
'/run': {}
},
Labels: {
"location": app.location,
"appId": app.id,
"isSubcontainer": String(!isAppContainer)
},
HostConfig: {
Binds: addons.getBindsSync(app, app.manifest.addons),
Memory: memoryLimit / 2,
MemorySwap: memoryLimit, // Memory + Swap
PortBindings: isAppContainer ? dockerPortBindings : { },
PublishAllPorts: false,
ReadonlyRootfs: !developmentMode, // see also Volumes in startContainer
RestartPolicy: {
"Name": isAppContainer ? "always" : "no",
"MaximumRetryCount": 0
},
CpuShares: 512, // relative to 1024 for system processes
VolumesFrom: isAppContainer ? null : [ app.containerId + ":rw" ],
NetworkMode: isolatedNetworkNs ? 'default' : ('container:' + app.containerId), // share network namespace with parent
Links: isolatedNetworkNs ? addons.getLinksSync(app, app.manifest.addons) : null, // links is redundant with --net=container
SecurityOpt: config.CLOUDRON ? [ "apparmor:docker-cloudron-app" ] : null // profile available only on cloudron
}
};
containerOptions = _.extend(containerOptions, options);
debugApp(app, 'Creating container for %s with options %j', app.manifest.dockerImage, containerOptions);
docker.createContainer(containerOptions, callback);
});
}
function createContainer(app, callback) {
createSubcontainer(app, app.id /* name */, null /* cmd */, { } /* options */, callback);
}
function startContainer(containerId, callback) {
assert.strictEqual(typeof containerId, 'string');
assert.strictEqual(typeof callback, 'function');
var docker = exports.connection;
var container = docker.getContainer(containerId);
debug('Starting container %s', containerId);
container.start(function (error) {
if (error && error.statusCode !== 304) return callback(new Error('Error starting container :' + error));
return callback(null);
});
}
function stopContainer(containerId, callback) {
assert(!containerId || typeof containerId === 'string');
assert.strictEqual(typeof callback, 'function');
if (!containerId) {
debug('No previous container to stop');
return callback();
}
var docker = exports.connection;
var container = docker.getContainer(containerId);
debug('Stopping container %s', containerId);
var options = {
t: 10 // wait for 10 seconds before killing it
};
container.stop(options, function (error) {
if (error && (error.statusCode !== 304 && error.statusCode !== 404)) return callback(new Error('Error stopping container:' + error));
debug('Waiting for container ' + containerId);
container.wait(function (error, data) {
if (error && (error.statusCode !== 304 && error.statusCode !== 404)) return callback(new Error('Error waiting on container:' + error));
debug('Container %s stopped with status code [%s]', containerId, data ? String(data.StatusCode) : '');
return callback(null);
});
});
}
function deleteContainer(containerId, callback) {
assert(!containerId || typeof containerId === 'string');
assert.strictEqual(typeof callback, 'function');
debug('deleting container %s', containerId);
if (containerId === null) return callback(null);
var docker = exports.connection;
var container = docker.getContainer(containerId);
var removeOptions = {
force: true, // kill container if it's running
v: true // removes volumes associated with the container (but not host mounts)
};
container.remove(removeOptions, function (error) {
if (error && error.statusCode === 404) return callback(null);
if (error) debug('Error removing container %s : %j', containerId, error);
callback(error);
});
}
function deleteContainers(appId, callback) {
assert.strictEqual(typeof appId, 'string');
assert.strictEqual(typeof callback, 'function');
var docker = exports.connection;
debug('deleting containers of %s', appId);
docker.listContainers({ all: 1, filters: JSON.stringify({ label: [ 'appId=' + appId ] }) }, function (error, containers) {
if (error) return callback(error);
async.eachSeries(containers, function (container, iteratorDone) {
deleteContainer(container.Id, iteratorDone);
}, callback);
});
}
function stopContainers(appId, callback) {
assert.strictEqual(typeof appId, 'string');
assert.strictEqual(typeof callback, 'function');
var docker = exports.connection;
debug('stopping containers of %s', appId);
docker.listContainers({ all: 1, filters: JSON.stringify({ label: [ 'appId=' + appId ] }) }, function (error, containers) {
if (error) return callback(error);
async.eachSeries(containers, function (container, iteratorDone) {
stopContainer(container.Id, iteratorDone);
}, callback);
});
}
function deleteImage(manifest, callback) {
assert(!manifest || typeof manifest === 'object');
assert.strictEqual(typeof callback, 'function');
var dockerImage = manifest ? manifest.dockerImage : null;
if (!dockerImage) return callback(null);
var docker = exports.connection;
var removeOptions = {
force: false, // might be shared with another instance of this app
noprune: false // delete untagged parents
};
// registry v1 used to pull down all *tags*. this meant that deleting image by tag was not enough (since that
// just removes the tag). we used to remove the image by id. this is not required anymore because aliases are
// not created anymore after https://github.com/docker/docker/pull/10571
docker.getImage(dockerImage).remove(removeOptions, function (error) {
if (error && error.statusCode === 404) return callback(null);
if (error && error.statusCode === 409) return callback(null); // another container using the image
if (error) debug('Error removing image %s : %j', dockerImage, error);
callback(error);
});
}