def main(argv=None):
parser = argparse.ArgumentParser(description="Convert Telemetry data")
parser.add_argument("-c",
"--config-file",
help="Read configuration from this file",
default="./telemetry_server_config.json")
parser.add_argument("-d",
"--date",
help="Use specified date for dimensions")
args = parser.parse_args()
try:
server_config = open(args.config_file, "r")
config = json.load(server_config)
server_config.close()
except IOError:
config = {}
cache_dir = config.get("revision_cache_path", "./histogram_cache")
server = config.get("revision_cache_server", "hg.mozilla.org")
schema_filename = config.get("schema_filename", "./telemetry_schema.json")
schema_data = open(schema_filename)
schema = TelemetrySchema(json.load(schema_data))
schema_data.close()
cache = revision_cache.RevisionCache(cache_dir, server)
converter = Converter(cache, schema)
process(converter, args.date)
def __init__(self, config):
# Sanity check args.
if config.num_mappers <= 0:
raise ValueError("Number of mappers must be greater than zero")
if config.num_reducers <= 0:
raise ValueError("Number of reducers must be greater than zero")
if not os.path.isdir(config.data_dir):
raise ValueError("Data dir must be a valid directory")
if not os.path.isdir(config.work_dir):
raise ValueError("Work dir must be a valid directory")
if not os.path.isfile(config.job_script):
raise ValueError("Job script must be a valid python file")
if not os.path.isfile(config.input_filter):
raise ValueError("Input filter must be a valid json file")
self._input_dir = config.data_dir
if self._input_dir[-1] == os.path.sep:
self._input_dir = self._input_dir[0:-1]
self._work_dir = config.work_dir
self._input_filter = TelemetrySchema(
json.load(open(config.input_filter)))
self._allowed_values = self._input_filter.sanitize_allowed_values()
self._output_file = config.output
self._num_mappers = config.num_mappers
self._num_reducers = config.num_reducers
self._local_only = config.local_only
self._bucket_name = config.bucket
self._aws_key = config.aws_key
self._aws_secret_key = config.aws_secret_key
modulefd = open(config.job_script)
## Lifted from FileDriver.py in jydoop.
self._job_module = imp.load_module("telemetry_job", modulefd,
config.job_script, ('.py', 'U', 1))
def setUpClass(cls):
cls.cache_dir = "/tmp/histogram_revision_cache"
cls.schema_filename = "./telemetry/telemetry_schema.json"
assert not os.path.exists(cls.cache_dir)
schema_file = open(cls.schema_filename, "r")
cls.schema = TelemetrySchema(json.load(schema_file))
schema_file.close()
cls.cache = revision_cache.RevisionCache(cls.cache_dir,
'hg.mozilla.org')
cls.converter = Converter(cls.cache, cls.schema)
def test_more_allowed(self):
spec = {
"version":
1,
"dimensions": [{
"field_name": "reason",
"allowed_values": ["saved-session"]
}, {
"field_name": "appName",
"allowed_values": "*"
}, {
"field_name": "appUpdateChannel",
"allowed_values": ["nightly"]
}, {
"field_name": "appVersion",
"allowed_values": "*"
}, {
"field_name": "appBuildID",
"allowed_values": "one_specific_build"
}, {
"field_name": "submission_date",
"allowed_values": {
"min": "20130908",
"max": "20140401"
}
}]
}
schema = TelemetrySchema(spec)
allowed = schema.sanitize_allowed_values()
self.assertTrue(schema.is_allowed("20130908", allowed[5]))
self.assertTrue(schema.is_allowed("20140401", allowed[5]))
self.assertTrue(schema.is_allowed("20130909", allowed[5]))
self.assertTrue(schema.is_allowed("20140101", allowed[5]))
self.assertFalse(schema.is_allowed("20130907", allowed[5]))
self.assertFalse(schema.is_allowed("20000000", allowed[5]))
self.assertFalse(schema.is_allowed("20140402", allowed[5]))
self.assertFalse(schema.is_allowed("99999999", allowed[5]))
self.assertTrue(schema.is_allowed("one_specific_build", allowed[4]))
self.assertFalse(schema.is_allowed("two_specific_build", allowed[4]))
self.assertFalse(schema.is_allowed("*", allowed[4]))
self.assertFalse(schema.is_allowed("one_specific_build ", allowed[4]))
self.assertFalse(schema.is_allowed("one-specific-build", allowed[4]))
def setUp(self):
self.schema = TelemetrySchema(self.get_schema_spec())
self.allowed_values = self.schema.sanitize_allowed_values()
def main():
parser = argparse.ArgumentParser(
description='Split raw logs into partitioned files.',
formatter_class=argparse.ArgumentDefaultsHelpFormatter)
parser.add_argument("-m",
"--max-output-size",
metavar="N",
help="Rotate output files after N bytes",
type=int,
default=500000000)
parser.add_argument("-i",
"--input-file",
help="Filename to read from",
required=True)
parser.add_argument("-o",
"--output-dir",
help="Base directory to store split files",
required=True)
parser.add_argument("-t",
"--telemetry-schema",
help="Filename of telemetry schema spec",
required=True)
parser.add_argument("-b", "--bucket", help="S3 Bucket name")
parser.add_argument("-k", "--aws-key", help="AWS Key")
parser.add_argument("-s", "--aws-secret-key", help="AWS Secret Key")
args = parser.parse_args()
schema_data = open(args.telemetry_schema)
schema = TelemetrySchema(json.load(schema_data))
schema_data.close()
storage = StorageLayout(schema, args.output_dir, args.max_output_size)
expected_dim_count = len(schema._dimensions)
if not os.path.isdir(args.output_dir):
os.makedirs(args.output_dir)
record_count = 0
fin = open(args.input_file, "rb")
bytes_read = 0
start = datetime.now()
while True:
record_count += 1
# Read two 4-byte values and one 8-byte value
lengths = fin.read(16)
if lengths == '':
break
len_path, len_data, timestamp = struct.unpack("<IIQ", lengths)
# Incoming timestamps are in milliseconds, so convert to POSIX first
# (ie. seconds)
submission_date = date.fromtimestamp(timestamp /
1000).strftime("%Y%m%d")
path = unicode(fin.read(len_path), errors="replace")
#print "Path for record", record_count, path, "length of data:", len_data
# Detect and handle gzipped data.
data = fin.read(len_data)
try:
# Note: from brief testing, cStringIO doesn't appear to be any
# faster. In fact, it seems slightly slower than StringIO.
data_reader = StringIO.StringIO(data)
uncompressor = gzip.GzipFile(fileobj=data_reader, mode="r")
data = unicode(uncompressor.read(), errors="replace")
uncompressor.close()
data_reader.close()
except Exception, e:
#print e
# Use the string as-is
data = unicode(data, errors="replace")
bytes_read += 8 + len_path + len_data
#print "Path for record", record_count, path, "length of data:", len_data, "data:", data[0:5] + "..."
path_components = path.split("/")
if len(path_components) != expected_dim_count:
# We're going to pop the ID off, but we'll also add the submission,
# so it evens out.
print "Found an invalid path in record", record_count, path
continue
key = path_components.pop(0)
info = {}
info["reason"] = path_components.pop(0)
info["appName"] = path_components.pop(0)
info["appVersion"] = path_components.pop(0)
info["appUpdateChannel"] = path_components.pop(0)
info["appBuildID"] = path_components.pop(0)
dimensions = schema.dimensions_from(info, submission_date)
#print " Converted path to filename", schema.get_filename(args.output_dir, dimensions)
storage.write(key, data, dimensions)
