quay/util/migrate/backfill_content_checksums.py

import logging

from peewee import JOIN_LEFT_OUTER

from peewee import (CharField, BigIntegerField, BooleanField, ForeignKeyField, DateTimeField,
                    TextField, fn)

from data.database import BaseModel, db, db_for_update, CloseForLongOperation
from app import app, storage
from digest import checksums
from util.migrate.allocator import yield_random_entries


logger = logging.getLogger(__name__)


class Repository(BaseModel):
  pass


# Vendor the information from tables we will be writing to at the time of this migration
class ImageStorage(BaseModel):
  uuid = CharField(index=True, unique=True)
  checksum = CharField(null=True)
  image_size = BigIntegerField(null=True)
  uncompressed_size = BigIntegerField(null=True)
  uploading = BooleanField(default=True, null=True)
  cas_path = BooleanField(default=True)
  content_checksum = CharField(null=True, index=True)


class Image(BaseModel):
  docker_image_id = CharField(index=True)
  repository = ForeignKeyField(Repository)
  ancestors = CharField(index=True, default='/', max_length=64535, null=True)
  storage = ForeignKeyField(ImageStorage, index=True, null=True)
  created = DateTimeField(null=True)
  comment = TextField(null=True)
  command = TextField(null=True)
  aggregate_size = BigIntegerField(null=True)
  v1_json_metadata = TextField(null=True)
  v1_checksum = CharField(null=True)


class ImageStorageLocation(BaseModel):
  name = CharField(unique=True, index=True)


class ImageStoragePlacement(BaseModel):
  storage = ForeignKeyField(ImageStorage)
  location = ForeignKeyField(ImageStorageLocation)



def _get_image_storage_locations(storage_id):
  placements_query = (ImageStoragePlacement
                      .select(ImageStoragePlacement, ImageStorageLocation)
                      .join(ImageStorageLocation)
                      .switch(ImageStoragePlacement)
                      .join(ImageStorage, JOIN_LEFT_OUTER)
                      .where(ImageStorage.id == storage_id))

  locations = set()
  for placement in placements_query:
    locations.add(placement.location.name)

  return locations


def backfill_content_checksums():
  """ Copies metadata from image storages to their images. """
  logger.debug('Image content checksum backfill: Began execution')

  def batch_query():
    return (ImageStorage
            .select(ImageStorage.id, ImageStorage.uuid)
            .where(ImageStorage.content_checksum >> None, ImageStorage.uploading == False))

  max_id = ImageStorage.select(fn.Max(ImageStorage.id)).scalar()

  for candidate_storage, abort in yield_random_entries(batch_query, ImageStorage.id, 1000, max_id):
    logger.debug('Computing content checksum for storage: %s', candidate_storage.uuid)

    locations = _get_image_storage_locations(candidate_storage.id)

    checksum = None
    with CloseForLongOperation(app.config):
      try:
        # Compute the checksum
        layer_path = storage.image_layer_path(candidate_storage.uuid)
        with storage.stream_read_file(locations, layer_path) as layer_data_handle:
          checksum = 'sha256:{0}'.format(checksums.sha256_file(layer_data_handle))
      except Exception as exc:
        logger.warning('Unable to compute checksum for storage: %s', candidate_storage.uuid)
        checksum = 'unknown:{0}'.format(exc.__class__.__name__)

    # Now update the ImageStorage with the checksum
    with app.config['DB_TRANSACTION_FACTORY'](db):
      to_update = db_for_update(ImageStorage.get(ImageStorage.id == candidate_storage.id))
      if to_update.content_checksum is not None:
        logger.info('Another worker filled in the checksum: %s', candidate_storage.uuid)
        abort.set()
      else:
        logger.debug('Setting content checksum to %s for %s', checksum, candidate_storage.uuid)
        to_update.content_checksum = checksum
        to_update.save()

if __name__ == '__main__':
  logging.basicConfig(level=logging.DEBUG)
  # logging.getLogger('peewee').setLevel(logging.CRITICAL)
  backfill_content_checksums()