Repository navigation
Performance Regression or Improvement: cogbk_python_batch_load_test_2GB_of_100B_records_with_a_single_key:runtime #40512
Description
Performance change found in the
test: cogbk_python_batch_load_test_2GB_of_100B_records_with_a_single_key for the metric: runtime.
For more information on how to triage the alerts, please look at
Triage performance alert issues section of the README.
Test description: CoGroupByKey Python Load test - 2GB of 100B records with a single key
python | coGBK | 100B records with a single key
Test Link - https://github.com/apache/beam/blob/5e38decf9e723a385057131b01bbd33d8c60bda3/.test-infra/jenkins/job_LoadTests_coGBK_Python.groovy#L32C25-L32C76
Test Dashboard - http://metrics.beam.apache.org/d/fK0U4JqWz/cogbk-load-tests?orgId=1&var-processingType=batch&var-sdk=python&viewPanel=2
timestamp: Sun Oct 11 15:45:35 2026, metric_value: 555.00
timestamp: Sat Oct 10 15:46:52 2026, metric_value: 457.00
timestamp: Fri Oct 9 15:48:38 2026, metric_value: 459.00 <---- Anomaly
timestamp: Thu Oct 8 15:47:38 2026, metric_value: 166.00
timestamp: Wed Oct 7 15:56:38 2026, metric_value: 287.00
timestamp: Tue Oct 6 15:41:26 2026, metric_value: 161.00
timestamp: Mon Oct 5 15:48:46 2026, metric_value: 449.00
timestamp: Sun Oct 4 17:06:17 2026, metric_value: 246.00
timestamp: Sat Oct 3 16:54:45 2026, metric_value: 247.00
timestamp: Fri Oct 2 15:49:05 2026, metric_value: 268.00
timestamp: Thu Oct 1 15:45:27 2026, metric_value: 278.00
timestamp: Wed Sep 30 15:50:54 2026, metric_value: 346.00
timestamp: Tue Sep 29 15:45:41 2026, metric_value: 232.00
Activity
Abacn commentedon Oct 12, 2026
Likely due to #40413, now running on Python3.11 on the last fast run (7a5f891)
Also note that it isn't overriding apache-beam-testing container image (it's using us-central1-artifactregistry.gcr.io/google.com/dataflow-containers/worker/v1beta3/beam_python3.11_sdk:beam-master-20260909). It should do therefore pending another fix
Update: Python 3.11 run at same commit (7a5f891) is also slow. Therefore this is pre-existing
Notice that Python3.11 SynthericSource seems to be much slower than Python3.10:
Python3.10:
Python3.11:
Probably the cause. Investigating the running time difference of syntheticsource...
Abacn commentedon Oct 13, 2026
Fixed by #40540