❤
npm
Sign UpSign In

cloud-compute-cannon-client

0.3.0 • Public • Published

Cloud Compute Cannon Build Status

Cloud Compute Cannon is a tool aimed at scientists and more general users who want to use cheap cloud providers (such as Amazon) to perform large scale computes (number crunching). It aims to lower some of the biggest barriers and learning curves in getting data and custom code running on distributed cloud infrastructure. It can be run both as a command-line tool, or as a server for integrating into other tools via a REST API/websockets.

Use cases:

  • Simulating molecular dynamics
  • Numerical simulations, data crunching
  • Server infrastructure for scalable computation

Cloud Compute Cannon is designed to do one thing well: run docker-based compute jobs on any cloud provider (or your local machine) reliably, with a minimum or user intervention, and scale machines up and down as needed. Its feature set is purposefully limited, it is designed to be used standalone, or as a component in more complex tools, rather than be extended itself.

Features:

  • Run on any cloud provider supported by pkgcloud or your local machine.
  • Adding providers is straightforward, so that developers can potentially add their own company or university cloud infrastructure.
  • One line installation.
  • Simple configuration.
  • Any programming language or tool can be used by jobs, as long as it runs under Linux (docker is used for running jobs).
  • Worker machines will scale up and down as needed.
  • The client can be run on the command line, or run as a server for providing an API to other tools.

API

Core computes:

/api/rpc/

Workflows

POST /api/rpc

{ "jsonrpc": "2.0", "method": "cwl", "params": { "git": "https://github.com/dionjwa/cwltool", "sha": "e7e6e18f62cf5db6541f15fedcee47ae0e219bbf", "cwl": "tests/ccc_docker_workflow/run_workflow.cwl", "input": "tests/ccc_docker_workflow/input.yml", "inputs": { } } }

Installation

Requirements

Install

Cloud-compute-cannon installs globally.

npm install -g cloud-compute-cannon

This installs the ccc executable.

Running

Environmental Variables

DISABLE_STARTUP_TEST

DISABLE_STARTUP_TEST=1

Default: false (tests will run by default)

To disable tests on startup:

DISABLE_STARTUP_TEST=0 | DISABLE_STARTUP_TEST=false

PORT

PORT=9000

COMPUTE_CONFIG

A string of the YAML config (below).

CONFIG_PATH

The path to the config file in the docker container.

Default: /app/config/ccc.yml

REMOVE_JOBS_ON_STARTUP

REMOVE_JOBS_ON_STARTUP=1 || REMOVE_JOBS_ON_STARTUP=true

Default: false

Removes all pending and running jobs on startup. Usually this is only useful for debugging or local dev environments.

LOG_LEVEL

LOG_LEVEL=20

10=trace 20=debug 30=info 40=warn 50=error 60=critical

Default: 20

To disable all logging:

DISABLE_LOGGING=true or DISABLE_LOGGING=1

ENABLE_FLUENT

ENABLE_FLUENT=false

Default: true

Allow disabling fluentd logging (on by default, and expecting the fluent daemon to be accessible with the host fluentd).

SCALE_UP_CONTROL

SCALE_UP_CONTROL=internal | external

Default: external

If SCALE_UP_CONTROL==internal then the stack can create new workers if the queue becomes larger than capacity. External means scaling is controlled via another process, e.g. AWS autoscaling groups.

SCALE_DOWN_CONTROL

SCALE_DOWN_CONTROL=internal | external

Default: external

If SCALE_DOWN_CONTROL==internal then when the local instance queue is empty the instance will terminate itself if it is cost effective to do so. E.g. AWS instances wait until just before a new billing interval (every hour). External means termination is controlled via another process, e.g. AWS autoscaling groups.

'Hello world' example

--command='["python", "-c", "print(\"Hello World!\")"]

ccc run --image=elyase/staticpython --command='["python", "-c", "print(\"Hello World!\")"]' --output=./localOutputDir

This assumes you have docker installed on the local machine, since we haven't specified any cloud provider, such as AWS. It instructs a worker to pull the python docker image, creates a container that runs the python command. The output option instructs cloudcannon to copy the results (here only the stdout) to the directory.

When the job is finished, you should see the text file ./localOutputDir/stdout containing the string "Hello world!".

There are many other options that are documented below.

Kibana/Elasticsearch Dashboard

You can see logs for the entire stack at:

http://:5601

The first time you'll need to configure Kibana:

Click on settings in the top-line menu, choose indices and then make sure that the index name contains 'logstash-*', then click in the 'time-field' name and choose '@timestamp';

To confirm there are indices go to:

http://:5601/_cat/indices

and you should see logstash-*** indices.

Utils

Get a docker-compose file to run a stack locally:

docker run --rm dionjwa/cloud-compute-cannon cat etc/docker-compose/single-server/docker-compose.yml > test.compose.yml

Configuration

Deploying to a remote machine, or deploying anywhere that uses real cloud providers (e.g. AWS, GCE) requires passing in a configuration file.

Example config yaml files are found in compute/servers/etc.

Yaml config example:

 
storage:
  type: "local"
  rootPath: "data/ServiceStorageLocalFileSystem"
 
providers:
  - type: "PkgCloud"
    maxWorkers: 2
    minWorkers: 0
    priority: 1
    #billingIncrement are measured in MINUTES. AWS defaults to 60 if this is not set 
    billingIncrement: 58
    credentials:
      provider: "amazon"
      keyId: "AKIAIWJON3HDJ"
      key: "SLV3lNIPbGwv6oSy"
      region: "us-west-1"
    server:
      Tags:
        - Key: "Name"
          Value: "CCC_Server"
      InstanceType: "m3.medium"
      ImageId: "ami-652c5505"
      KeyName: "ccc-keypair"
      Key: |
        -----BEGIN RSA PRIVATE KEY-----
        MIIEogIBAAKCAQEAyh+2Op9GIcRjlayC+TP6Btxklb9nkQlrKJaXlovJfHgQPOvpTnDceyzHy755
        JpsCbhrdio4GomeKHBObbD4eB5nIZ8VXQD1EhgedUxKKrW9csWyjlRbfOWEZyMmT025JIg8G4QYK
        -----END RSA PRIVATE KEY-----
    worker:
      Tags:
        - Key: "Name"
          Value: "CCC_Worker"
      InstanceType: "m3.medium"
      ImageId: "ami-61e99101"
      KeyName: "ccc-keypair"
      Key: |
        -----BEGIN RSA PRIVATE KEY-----
        MIIEogIBAAKCAQEAyh+2Op9GIcRjlayC+TP6Btxklb9nkQlrKJaXlovJfHgQPOvpTnDceyzHy755
        JpsCbhrdio4GomeKHBObbD4eB5nIZ8VXQD1EhgedUxKKrW9csWyjlRbfOWEZyMmT025JIg8G4QYK
        -----END RSA PRIVATE KEY-----

In the above configuration, only the credentials.keyId and credentials.key values need to be modified to run under your own AWS account.

Storage providers

  storage:
    type: "local"
    rootPath: "data/ServiceStorageLocalFileSystem"

  storage:
    type: "S3"
    rootPath: "/"
    container: "bucket-name"
    # You can use the S3 url or the cloudfront. The httpAccessUrl
    # is prepended to the file path.
    #httpAccessUrl: "https://s3-us-west-1.amazonaws.com/bucket-name"
    httpAccessUrl: "https://d3bt947tva2i1l.cloudfront.net"
    credentials:
      provider: "amazon"
      #You can use keyId/key or accessKeyId/secretAccessKey field names
      keyId: "your keyId"
      key: "your key"
      region: "us-west-1"
    # Optionally add arbitrary  copy parameters if using S3:
    # http://docs.aws.amazon.com/cli/latest/reference/s3/sync.html
    extraS3SyncParameters: [
        ['--exclude', '*', '--include', '*.ext2'],
        ['--exclude', '*', '--include', '*.ext3']
      ]

More coming soon!

Testing

Call the this API endpoint to run the suite of internal functional end-to-end tests:

curl HOST:PORT/api/rpc/server-tests

This will return the tests results in JSON and a status of 200 if the tests pass, or a 500 status if any test fails.

Developers

CloudComputeCannon is written in Haxe. To develop, you'll need to install Haxe, (Docker +1.12, and Node+NPM.

Steps for running the functional tests locally:

  1. Clone the repo and install libraries, submodules, etc

    git clone https://github.com/Autodesk/cloud-compute-cannon cd cloud-compute-cannon npm run init

The npm run init command wraps steps that are identical to the Dockerfile, since that creates an environment for compiling and running the server.

To compile, run:

haxe etc/hxml/build-all.hxml

To run you'll need the rest of the stack (database, registry, logging). In a new terminal window in the cloud-compute-cannon directory run the following command. It assumes you have (Docker installed:

./bin/run-stack-noserver

Then in another terminal window, run:

haxe test/testsIntegration.hxml

To run individual tests:

./bin/test <full.class.name>[.optionalMethod]

Running tests on cloud providers

  1. Set up an instance on e.g. AWS. Map the IP address to an entry in your ~/.ssh/config for passwordless operations. We'll call our alias "dev".

  2. Create your server configuration file (default name="ccc.yml", see above).

./bin/reloading-stack-deploy <ssh alias (dev)>

  1. On server compiles

./bin/reloading-stack-sync <ssh alias (dev)>

The server code and config will be copied over and the server restarted. Restaring means running the tests.

Optionally create an .env file in the root of the repo, this will be consumed by the reloading server.

Developing locally and having compiled server builds automatically uploaded to a remote server and restarted

  1. Set up an instance on e.g. AWS. Map the IP address to an entry in your ~/.ssh/config for passwordless operations. We'll call our alias "dev".

./bin/run-stack-remote-dev <ccc.yml>

  1. Code and build! Updates are automatically pushed to the remote server!

Contact

To contact the authors, please email: maintainers.bionano-cloudcomputecannon@autodesk.com or post an issue.

License

Copyright 2015 Autodesk Inc.

Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with the License. You may obtain a copy of the License at

http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License.

Readme

Keywords

none

Package Sidebar

Install

npm i cloud-compute-cannon-client

Repository

github.com/Autodesk/cloud-compute-cannon

Homepage

github.com/Autodesk/cloud-compute-cannon#readme

Weekly Downloads

2

Version

0.3.0

License

none

Last publish

Collaborators

  • dionjwa
Try on RunKit
Report malware