quay/buildman/jobutil/buildjob.py

import json
import logging

from cachetools import lru_cache
from endpoints.notificationhelper import spawn_notification
from data import model
from util.imagetree import ImageTree
from util.morecollections import AttrDict

logger = logging.getLogger(__name__)


class BuildJobLoadException(Exception):
  """ Exception raised if a build job could not be instantiated for some reason. """
  pass

class BuildJob(object):
  """ Represents a single in-progress build job. """
  def __init__(self, job_item):
    self.job_item = job_item

    try:
      self.job_details = json.loads(job_item.body)
    except ValueError:
      raise BuildJobLoadException(
          'Could not parse build queue item config with ID %s' % self.job_details['build_uuid']
      )

  @property
  def retries_remaining(self):
    return self.job_item.retries_remaining

  def has_retries_remaining(self):
    return self.job_item.retries_remaining > 0

  def send_notification(self, kind, error_message=None, image_id=None, manifest_digests=None):
    tags = self.build_config.get('docker_tags', ['latest'])
    event_data = {
      'build_id': self.repo_build.uuid,
      'build_name': self.repo_build.display_name,
      'docker_tags': tags,
      'trigger_id': self.repo_build.trigger.uuid,
      'trigger_kind': self.repo_build.trigger.service.name,
      'trigger_metadata': self.build_config.get('trigger_metadata', {})
    }

    if image_id is not None:
      event_data['image_id'] = image_id

    if manifest_digests:
      event_data['manifest_digests'] = manifest_digests

    if error_message is not None:
      event_data['error_message'] = error_message

    # TODO(jzelinskie): remove when more endpoints have been converted to using
    # interfaces
    repo = AttrDict({
      'namespace_name': self.repo_build.repository.namespace_user.username,
      'name': self.repo_build.repository.name,
    })
    spawn_notification(repo, kind, event_data,
                       subpage='build/%s' % self.repo_build.uuid,
                       pathargs=['build', self.repo_build.uuid])


  @lru_cache(maxsize=1)
  def _load_repo_build(self):
    try:
      return model.build.get_repository_build(self.build_uuid)
    except model.InvalidRepositoryBuildException:
      raise BuildJobLoadException(
          'Could not load repository build with ID %s' % self.build_uuid)

  @property
  def build_uuid(self):
    """ Returns the unique UUID for this build job. """
    return self.job_details['build_uuid']

  @property
  def namespace(self):
    """ Returns the namespace under which this build is running. """
    return self.repo_build.repository.namespace_user.username

  @property
  def repo_name(self):
    """ Returns the name of the repository under which this build is running. """
    return self.repo_build.repository.name

  @property
  def repo_build(self):
    return self._load_repo_build()

  def get_build_package_url(self, user_files):
    """ Returns the URL of the build package for this build, if any or empty string if none. """
    archive_url = self.build_config.get('archive_url', None)
    if archive_url:
      return archive_url

    if not self.repo_build.resource_key:
      return ''

    return user_files.get_file_url(self.repo_build.resource_key, requires_cors=False)

  @property
  def pull_credentials(self):
    """ Returns the pull credentials for this job, or None if none. """
    return self.job_details.get('pull_credentials')

  @property
  def build_config(self):
    try:
      return json.loads(self.repo_build.job_config)
    except ValueError:
      raise BuildJobLoadException(
        'Could not parse repository build job config with ID %s' % self.job_details['build_uuid']
      )

  def determine_cached_tag(self, base_image_id=None, cache_comments=None):
    """ Returns the tag to pull to prime the cache or None if none. """
    cached_tag = None
    if base_image_id and cache_comments:
      cached_tag = self._determine_cached_tag_by_comments(base_image_id, cache_comments)

    if not cached_tag:
      cached_tag = self._determine_cached_tag_by_tag()

    logger.debug('Determined cached tag %s for %s: %s', cached_tag, base_image_id, cache_comments)

    return cached_tag

  def _determine_cached_tag_by_comments(self, base_image_id, cache_commands):
    """ Determines the tag to use for priming the cache for this build job, by matching commands
        starting at the given base_image_id. This mimics the Docker cache checking, so it should,
        in theory, provide "perfect" caching.
    """
    # Lookup the base image in the repository. If it doesn't exist, nothing more to do.
    repo_build = self.repo_build
    repo_namespace = repo_build.repository.namespace_user.username
    repo_name = repo_build.repository.name

    base_image = model.image.get_image(repo_build.repository, base_image_id)
    if base_image is None:
      return None

    # Build an in-memory tree of the full heirarchy of images in the repository.
    all_images = model.image.get_repository_images_without_placements(repo_build.repository,
                                                                      with_ancestor=base_image)

    all_tags = model.tag.list_repository_tags(repo_namespace, repo_name)
    tree = ImageTree(all_images, all_tags, base_filter=base_image.id)

    # Find a path in the tree, starting at the base image, that matches the cache comments
    # or some subset thereof.
    def checker(step, image):
      if step >= len(cache_commands):
        return False

      full_command = '["/bin/sh", "-c", "%s"]' % cache_commands[step]
      logger.debug('Checking step #%s: %s, %s == %s', step, image.id, image.command, full_command)

      return image.command == full_command

    path = tree.find_longest_path(base_image.id, checker)
    if not path:
      return None

    # Find any tag associated with the last image in the path.
    return tree.tag_containing_image(path[-1])


  def _determine_cached_tag_by_tag(self):
    """ Determines the cached tag by looking for one of the tags being built, and seeing if it
        exists in the repository. This is a fallback for when no comment information is available.
    """
    tags = self.build_config.get('docker_tags', ['latest'])
    repository = self.repo_build.repository
    existing_tags = model.tag.list_repository_tags(repository.namespace_user.username,
                                                   repository.name)
    cached_tags = set(tags) & set([tag.name for tag in existing_tags])
    if cached_tags:
      return list(cached_tags)[0]

    return None