An on-demand deployment web service for StackViz.

• Free software: Apache license
• Documentation: http://docs.openstack.org/developer/stackviz-deployer
• Source: http://git.openstack.org/cgit/openstack/stackviz-deployer
• Bugs: http://bugs.launchpad.net/stackviz

This project aims to provide a easy-to-use UI for generating StackViz sites. Users should be able to paste in a URL, select a list of matches, and then have a StackViz site generated on demand. The actual contents of the generated site should depend on what artifacts can be scraped from the job output (e.g. subunit, dstat, possibly more in the future).

This project is currently working towards a working proof-of-concept. Components are making varying degrees of progress toward this:

• API: mostly feature-complete, still needs input validation and cleanup jobs
• Workers: mostly feature-complete, still needs input validation
• Frontend: basic implementation is finished, see public/
• StackViz support: done, see

The server implementation is reasonably complete and should be fine for simple deployments. Right now it uses MySQL for persistent storage with a Celery task queue (using Redis as a broker). In production the web UI (static HTML) and API server (python+flask) should be run behind Nginx or Apache.

The current requirements are as follows:

• MySQL server on localhost, with empty database stackviz, user stackviz, password stackviz
• Redis server on localhost, default port, no authentication (will use db #0)
• A celery worker: celery -A stackviz_deployer.tasks.tasks worker
• The API server: PYTHONPATH="." python -mstackviz_deployer.api.api

Several environment variables are also available to override defaults (using naming conventions for linked Docker containers):

• MYSQL_ENV_MYSQL_USER: MySQL user, default 'stackviz'
• MYSQL_ENV_MYSQL_PASSWORD: MySQL password, default 'stackviz'
• MYSQL_ENV_MYSQL_DATABASE: MySQL database, default 'stackviz'
• MYSQL_PORT_3306_TCP_ADDR: MySQL host address, default 'localhost'
• MYSQL_PORT_3306_TCP_PORT: MySQL port, default '3306'

Note that the MySQL database could get large relatively fast as gzipped artifacts are stored as blobs for the moment. That said, there should be no harm in purging records after some relatively short time limit (e.g. 7 days). Even so, one processed dataset (gzipped in the database, without logging) should be around 250 KB.

For production deployments, Nginx or Apache should be configured with the following routes (using nginx location syntax):

• / -> ./public/
• /api -> stackviz_deployer (via uwsgi, mod_python, etc)
• ^/s/[A-Za-z0-9\-]+/(.*)$ -> ./stackviz/build/$1
• /go/* -> ./public/go.html
  • All non-file addresses should be rewritten to go.html using try_files or similar

For a working example, see the Docker example using nginx and uwsgi, or the dev nginx config at etc/nginx-dev.conf.

The frontend is a plain HTML and JS site which can be found under the public/ directory. During development, it can be served using any HTTP server, such as twistd or Python's built-in SimpleHTTPServer.

Note that the frontend does require special handling of the /go/ route which may complicate testing in development, see the dev server instructions below for tips.

StackViz itself is also required, along with a special configuration file.

1. A production build should be performed using gulp prod
2. The deployer config should be added to stackviz/build/data/ (see etc/config.json)

The output, stackviz/build/, should then be served as plain static files from the /s/[uuid]/* route. During development, the normal StackViz dev server (gulp dev) is generally appropriate, assuming the correct config.json is added instead to app/data/ and the dev server is configured to proxy it as needed.

A small development server can be used to serve the frontend and proxy the API server to simulate a production environment. To use, first run the dev API server and then:

sudo npm install -g fe-dev-server
fds

The server will start on port 5001 an will serve a proxied API server at http://localhost:5001/api. The frontend should be available by browsing to http://localhost:5001/.

Assuming the normal StackViz dev server is running on its default port (3000), it will also be proxied so requests to /s/[uuid]/ will load StackViz as expected. If all goes well, you should have a fully-functional development environment.

Alternatively, an example nginx configuration suitable for development can be found at etc/nginx-dev.conf. This configuration handles routing between the various development servers and includes basic instructions for use in the file header.

The deployer supports additional log types and parsers that StackViz proper can't support, particularly those that make more expensive network requests or need additional datasets. These are generally optional, but helpful.

Using JJB definitions, the console.html parser can split and annotate console logs with the actual script names from project-config's YAML definitions. To load these definitions, either:

• Clone project-config to the working directory for the celery workers (presumably just stackviz-deployer/), or
• Set the environment variable JJB_YAML_PATH for celery workers (should be /path/to/project-config/jenkins/jobs)

Examples using HTTPie and jq with the dev server:

• List the latest Jenkins results for a Gerrit change (one entry per job):

  $ http post localhost:5000/list q=271726
  HTTP/1.0 200 OK
  Content-Length: 2379
  Content-Type: application/json
  Date: Tue, 09 Feb 2016 02:39:10 GMT
  Server: Werkzeug/0.10.4 Python/2.7.8
  
  { "results": [
      {
          "change_id": 271726,
          "change_project": "openstack/stackviz",
          "change_subject": "Add nprogress progress bars to timeline and test-details.",
          "ci_username": "jenkins",
          "name": "gate-stackviz-pep8",
          "pipeline": "check",
          "revision": 2,
          "status": "SUCCESS",
          "url": "http://logs.openstack.org/26/271726/2/check/gate-stackviz-pep8/7c374a7/"
      },
  ], ... }

• List Jenkins jobs for a specific Gerrit revision:

  $ http post localhost:5000/list q=271726,1

• List jobs from a Gerrit URL (with or without a revision or fragment):

  $ http post localhost:5000/list q='https://review.openstack.org/#/c/271726/'

• List a job directly from the artifact URL (will parse and look up Gerrit details when possible):

  $ http post localhost:5000/list q='http://logs.openstack.org/26/271726/2/gate/gate-stackviz-python27/937cf7b/'
  HTTP/1.0 200 OK
  Content-Length: 423
  Content-Type: application/json
  Date: Tue, 09 Feb 2016 02:45:30 GMT
  Server: Werkzeug/0.10.4 Python/2.7.8
  {
      "results": [
          {
              "change_id": 271726,
              "change_project": "openstack/stackviz",
              "change_subject": "Add nprogress progress bars to timeline and test-details.",
              "ci_username": null,
              "name": "gate-stackviz-python27",
              "pipeline": "gate",
              "revision": 2,
              "status": "SUCCESS",
              "url": "http://logs.openstack.org/26/271726/2/gate/gate-stackviz-python27/937cf7b/"
          }
      ]
  }

• Request a scrape of some artifact listing from /list:

  $ http post localhost:5000/list q=269624 | jq '.results[2]' | http post localhost:5000/scrape
  HTTP/1.0 202 ACCEPTED
  Content-Length: 74
  Content-Type: application/json
  Date: Tue, 09 Feb 2016 03:33:23 GMT
  Server: Werkzeug/0.10.4 Python/2.7.8
  
  {
      "status": "queued",
      "uuid": "f223e63b-6ac0-4236-9c1c-4dec769310aa"
  }

• Get the status of a scrape:

  $ http post localhost:5000/status q=f223e63b-6ac0-4236-9c1c-4dec769310aa
  HTTP/1.0 200 OK
  Content-Length: 76
  Content-Type: application/json
  Date: Tue, 09 Feb 2016 03:34:44 GMT
  Server: Werkzeug/0.10.4 Python/2.7.8
  
  {
      "status": "finished",
      "uuid": "f223e63b-6ac0-4236-9c1c-4dec769310aa"
  }

• Get the results of a scrape:

  http post localhost:5000/task q=f223e63b-6ac0-4236-9c1c-4dec769310aa
  HTTP/1.0 200 OK
  Content-Length: 761
  Content-Type: application/json
  Date: Tue, 09 Feb 2016 03:35:39 GMT
  Server: Werkzeug/0.10.4 Python/2.7.8
  
  {
      "artifacts": [
          {
              "artifact_type": "dstat",
              "content_encoding": "gzip",
              "content_type": "text/csv",
              "id": "09890181-4149-4cb2-82e3-c27f8301db03"
          },
          {
              "artifact_type": "subunit",
              "content_encoding": "gzip",
              "content_type": "application/json",
              "id": "79f81039-d51c-46af-a8cb-13e31efe1a57"
          }
      ],
      "change_id": 269624,
      "change_project": "openstack/cinder",
      "change_subject": "Support for consistency groups in ScaleIO driver",
      "ci_username": "jenkins",
      "id": "f223e63b-6ac0-4236-9c1c-4dec769310aa",
      "name": "gate-tempest-dsvm-full",
      "pipeline": "check",
      "revision": 19,
      "status": "SUCCESS",
      "url": "http://logs.openstack.org/24/269624/19/check/gate-tempest-dsvm-full/84f9b4a/"
  }

• Fetch an artifact blob (will have encoding and content type set appropriately):

  $ http get localhost:5000/blob/09890181-4149-4cb2-82e3-c27f8301db03 --headers
  HTTP/1.0 200 OK
  Content-Encoding: gzip
  Content-Length: 187744
  Content-Type: text/csv
  Date: Tue, 09 Feb 2016 03:36:57 GMT
  Server: Werkzeug/0.10.4 Python/2.7.8

Note that all API endpoints accept and produce JSON, except /blob.