def extend(d): try: db.insert_list(table_name, d) db.flush() Log.note("added {{num}} records", {"num":len(d)}) except Exception, e: Log.warning("Can not inert into database", e)
def extract_from_datazilla_using_id(es, settings, transformer): existing_ids = get_existing_ids(es, settings, transformer.pushlog.keys()) max_existing_id = nvl(MAX(existing_ids), settings.production.min) holes = set(range(settings.production.min, max_existing_id)) - existing_ids missing_ids = set(range(settings.production.min, max_existing_id+nvl(settings.production.step, NUM_PER_BATCH))) - existing_ids Log.note("Number missing: {{num}}", {"num": len(missing_ids)}) Log.note("Number in holes: {{num}}", {"num": len(holes)}) #FASTER IF NO INDEXING IS ON es.set_refresh_interval(-1) #FILE IS FASTER THAN NETWORK if (len(holes) > 10000 or settings.args.scan_file or settings.args.restart) and File(settings.param.output_file).exists: #ASYNCH PUSH TO ES IN BLOCKS OF 1000 with Timer("Scan file for missing ids"): with ThreadedQueue(es, size=nvl(es.settings.batch_size, 100)) as json_for_es: num = 0 for line in File(settings.param.output_file): try: if len(line.strip()) == 0: continue col = line.split("\t") id = int(col[0]) # if id==3003529: # Log.debug() if id < settings.production.min: continue if id in existing_ids: continue if num > settings.production.step: return num += 1 with Profiler("decode and transform"): data = CNV.JSON2object(col[-1]) if data.test_run_id: with Profiler("transform"): data = transformer.transform(id, data) json_for_es.extend({"value": d} for d in data) Log.note("Added {{id}} from file", {"id": id}) existing_ids.add(id) else: Log.note("Skipped {{id}} from file (no test_run_id)", {"id": id}) num -= 1 except Exception, e: Log.warning("Bad line id={{id}} ({{length}}bytes):\n\t{{prefix}}", { "id": id, "length": len(CNV.object2JSON(line)), "prefix": CNV.object2JSON(line)[0:130] }, e) missing_ids = missing_ids - existing_ids
def arrays_add(id, path, r): try: if isinstance(r, dict): for k, v in [(k, v) for k, v in r.items()]: new_path = path + "[" + k + "]" arrays_add(id, new_path, v) elif isinstance(r, list): try: values = r.map(float) arrays.append([id, path, len(values), 1]) except Exception, e: for i, v in enumerate(r): r[i] = arrays_add(id, path + "[" + str(i) + "]", v) # return r except Exception, e: Log.warning("Can not summarize: {{json}}", {"json": CNV.object2JSON(r)})
def etl(es_sink, file_sink, settings, transformer, max_id, id): """ PULL FROM DZ AND PUSH TO es AND file_sink """ # DEBUG GROWTH # with GC_LOCKER: # try: # if COUNTER.count % 100 == 0: # # gc.collect() # deltas, stats = objgraph.get_growth() # Log.note("Deltas:\n{{deltas|indent}}", {"deltas": deltas}) # except Exception, e: # Log.warning("objgraph problem", e) # # COUNTER.count += 1 url = settings.production.blob_url + "/" + str(id) try: with Timer("read {{id}} from DZ", {"id": id}): content = requests.get(url, timeout=nvl(settings.production.timeout, 30)).content except Exception, e: Log.warning("Failure to read from {{url}}", {"url": url}, e) return False
def transform(self, id, datazilla): try: r = datazilla.json_blob #ADD DATAZILLA MARKUP r.datazilla = { "id": id, "date_loaded": datazilla.date_loaded * 1000, "error_flag": datazilla.error_flag, "test_run_id": datazilla.test_run_id, "processed_flag": datazilla.processed_flag, "error_msg": datazilla.error_msg } #CONVERT UNIX TIMESTAMP TO MILLISECOND TIMESTAMP r.testrun.date *= 1000 def mainthread_transform(r): if r == None: return None output = Struct() for i in r.mainthread_readbytes: output[literal_field(i[1])].name = i[1] output[literal_field(i[1])].readbytes = i[0] r.mainthread_readbytes = None for i in r.mainthread_writebytes: output[literal_field(i[1])].name = i[1] output[literal_field(i[1])].writebytes = i[0] r.mainthread_writebytes = None for i in r.mainthread_readcount: output[literal_field(i[1])].name = i[1] output[literal_field(i[1])].readcount = i[0] r.mainthread_readcount = None for i in r.mainthread_writecount: output[literal_field(i[1])].name = i[1] output[literal_field(i[1])].writecount = i[0] r.mainthread_writecount = None r.mainthread = output.values() mainthread_transform(r.results_aux) mainthread_transform(r.results_xperf) #ADD PUSH LOG INFO try: branch = r.test_build.branch if branch.endswith("-Non-PGO"): r.test_build.branch = branch r.test_build.pgo = False branch = branch[0:-8] else: r.test_build.pgo = True with Profiler("get from pushlog"): if not self.pushlog: #NO PUSHLOG MEANS WE DO NOTHING TO MARKUP TEST RESULTS pass elif self.pushlog[branch]: possible_dates = self.pushlog[branch][r.test_build.revision] if possible_dates: r.test_build.push_date = int(Math.round(possible_dates[0].date * 1000)) else: if r.test_build.revision == 'NULL': r.test_build.no_pushlog = True # OOPS! SOMETHING BROKE elif CNV.milli2datetime(Math.min(r.testrun.date, r.datazilla.date_loaded)) < PUSHLOG_TOO_OLD: Log.note("{{branch}} @ {{revision}} has no pushlog, transforming anyway", r.test_build) r.test_build.no_pushlog = True else: Log.note("{{branch}} @ {{revision}} has no pushlog, try again later", r.test_build) return [] # TRY AGAIN LATER else: with self.locker: if branch not in self.unknown_branches: Log.note("Whole branch {{branch}} has no pushlog", {"branch":branch}) self.unknown_branches.add(branch) if CNV.milli2datetime(Math.min(r.testrun.date, r.datazilla.date_loaded)) < PUSHLOG_TOO_OLD: r.test_build.no_pushlog = True else: r.test_build.no_pushlog = True #return [r] #TODO: DO THIS IF WE FIGURE OUT HOW TO HANDLE THE VERY LARGE NUMBER OF RESULTS WITH NO PUSHLOG except Exception, e: Log.warning("{{branch}} @ {{revision}} has no pushlog", r.test_build, e) new_records = [] # RECORD THE UNKNOWN PART OF THE TEST RESULTS remainder = r.copy() remainder.results = None if len(remainder.keys()) > 4: new_records.append(remainder) #RECORD TEST RESULTS total = StructList() if r.testrun.suite in ["dromaeo_css", "dromaeo_dom"]: #dromaeo IS SPECIAL, REPLICATES ARE IN SETS OF FIVE #RECORD ALL RESULTS for i, (test_name, replicates) in enumerate(r.results.items()): for g, sub_results in Q.groupby(replicates, size=5): new_record = Struct( test_machine=r.test_machine, datazilla=r.datazilla, testrun=r.testrun, test_build=r.test_build, result={ "test_name": unicode(test_name) + "." + unicode(g), "ordering": i, "samples": sub_results } ) try: s = stats(sub_results) new_record.result.stats = s total.append(s) except Exception, e: Log.warning("can not reduce series to moments", e) new_records.append(new_record)
test_machine=r.test_machine, datazilla=r.datazilla, testrun=r.testrun, test_build=r.test_build, result={ "test_name": test_name, "ordering": i, "samples": replicates } ) try: s = stats(replicates) new_record.result.stats = s total.append(s) except Exception, e: Log.warning("can not reduce series to moments", e) new_records.append(new_record) if len(total) > 1: # ADD RECORD FOR GEOMETRIC MEAN SUMMARY new_record = Struct( test_machine=r.test_machine, datazilla=r.datazilla, testrun=r.testrun, test_build=r.test_build, result={ "test_name": "SUMMARY", "ordering": -1, "stats": geo_mean(total) }
added.add(id) data = CNV.JSON2object(col[1]) records_for_db.add({ "id": nvl(data.test_run_id, id), "branch": data.json_blob.test_build.branch, "name": data.json_blob.test_build.name, "version": data.json_blob.test_build.version, "suite": data.json_blob.testrun.suite, "revision": data.json_blob.test_build.revision, "date": data.json_blob.testrun.date }) Log.note("Added {{id}} from file", {"id": data.test_run_id}) except Exception, e: Log.warning("Bad line ({{length}}bytes):\n\t{{prefix}}", { "length": len(CNV.object2JSON(line)), "prefix": CNV.object2JSON(line)[0:130] }, e) def main(): try: settings = startup.read_settings(filename="file2db_settings.json") Log.start(settings.debug) with DB(settings.db) as db: db.execute(""" DROP TABLE IF EXISTS b2g_tests """) db.execute("""
date = CNV.unix2datetime(data.testrun.date) if id % 1000 == 0: Log.println("loading id " + str(id) + " date: " + CNV.datetime2string(date, "%Y-%m-%d %H:%M:%S")) if date < MINIMUM_DATE: continue if id in all: continue all.add(id) arrays_add(id, "[" + data.test_build.branch + "][" + data.testrun.suite + "]", data) output_file.write(str(id) + "\t" + json) except Exception, e: Log.warning("can not process line:\n\t" + line, e) smallest = min(*all) Log.println("First id >= date: {{min}}", {"min": smallest}) df = DataFrame(arrays, columns=["id", "path", "length", "count"]) colNames = [str(p) + " to " + str(parts[i + 1] - 1) for i, p in enumerate(parts[0:-1])] # http://pandas.pydata.org/pandas-docs/stable/groupby.html#na-group-handling length_dim = pandas.cut(df.length, parts, labels=colNames, right=False) summary = df.groupby(["path", length_dim], sort=False).size() #summary=summary.reindex(length_dim, level="length") table = summary.unstack("length") s = CNV.DataFrame2string(table)#, columns=colNames) Log.println("\n" + s) with open("talos_big_array_summary.tab", "w") as output_file:
es_sink.extend({"value": d} for d in result) file_sink.add(str(id) + "\t" + content + "\n") elif data.error_flag == 'Y': error = data.json_blob error.datazilla = data error.results = None data.json_blob = None es_sink.add({"value": error}) else: Log.println("No test run id for {{id}}", {"id": id}) del data return True except Exception, e: Log.warning("Failure to etl (content length={{length}})", {"length": len(content)}, e) return False def get_existing_ids(es, settings, branches): #FIND WHAT'S IN ES bad_ids = [] int_ids = set() demand_pushlog = {"match_all":{}} if branches: demand_pushlog = {"or": [ {"not": {"missing": {"field": "test_build.push_date"}}}, {"not": {"missing": {"field": "test_build.no_pushlog"}}} ]}