It's a REAL pain to figure out how to grab the data from Google cloud logging. This script lets you grab historic data and save locally so you can process since Cloud Logging Sync's won't do anything historic.

fetch_from_cloud_logging.sh

gcloud beta logging read "resource.type=\"dataflow_step\" resource.labels.job_id=\"2018-11-13_09_13_59-6912497806535022683\" OR \"timestamp>=\\\"2018-11-12T00:00:00Z\" OR \"timing:\" timestamp<=\"2200-01-01T00:00:00.000000000Z\" timestamp<\"2018-11-20T20:20:01.065Z\"" --format=json --limit 100000 | jq -rnc --stream 'fromstream(1|truncate_stream(inputs)) | .jsonPayload.message' >> gcs.txt

stat.py

filename = "gcs.txt"
file = open(filename, "r")
data = file.read()

data = data.replace('\t', '').split("\n---> Method timing:")
data = filter(None, data)  # fastest

results = {}
for item in data:

    item = item.strip().split('\n')
    for i in item:
        i = i.split(': ')

        keys = ['TOTAL',
                'apply_routing',
                'apply_scrubbing',
                'calculate_consumption',
                'calculate_pricing',
                'apply_leg_segments',
                'write_pricing']

        if i[0] not in keys:
            continue

        if i[0] not in results:
            results[i[0]] = []
        results[i[0]].append(float(i[1]))
    # print(item)

final_results = {}

for key, value in results.iteritems():
    final_results[key] = round(reduce(lambda x, y: x + y, value) / len(value), 2)

final_results['TOTAL_WO_WRITING'] = round(final_results['TOTAL'] - final_results['write_pricing'], 2)

from pprint import pprint

pprint(final_results)

print(len(results['TOTAL']))