Dataflow Streaming Pipeline Update Test

Description Test code for low latency streaming pipeline updates

Features

Getting Started

1. Clone this repo and switch to the checked out directory
2. Designate or create a project to run the tests and create terraform/terraform.tfvars file with the following content:

project_id = "<your project id>"

3. Create infrastructure to run:

cd terraform
terraform init
terraform apply
cd ..

Running Tests

Start the test harness

This will start a pipeline which will be generating synthetic events:

./start-event-generation.sh <rate>

A typical rate is tens of thousands of events per second. A Dataflow pipeline named data-generator-<rate> will be started. You can simulate event load increases by starting additional pipelines. Note, that you can't start several pipelines with exactly the same rate because the pipeline name needs to be unique.

You can see the current publishing load by summing the rates of all active data generation pipelines.

Start the baseline consumption pipeline

This pipeline runs uninterrupted on a dedicated PubSub subscription. The goals are to collect the message processing latencies under the perfect circumstances and the unique message ids in order to later compare them with the pipelines being tested.

./start-baseline-pipeline.sh

Start the pipeline which will be updated

./start-main-pipeline.sh

Update the pipeline

We are going to use the same pipeline code to update the existing pipeline - there is no difference in processing time

./update-pipeline.sh

Analyse the data

All scripts below have time ranges defined in the beginning of the scripts, typically 20 minute intervals. Adjust them as needed. For example, if you would like to check for a fixed time period add these lines after DECLARE statements:

SET start_ts = TIMESTAMP '2023-06-06 08:32:00.00 America/Los_Angeles';
SET end_ts = TIMESTAMP '2023-06-06 08:45:00.00 America/Los_Angeles';

Event latencies comparison

Use the following query to compare latency of ingestion of the baseline pipeline and the main pipeline:

DECLARE
  start_ts DEFAULT TIMESTAMP_ADD(CURRENT_TIMESTAMP(), INTERVAL -40 MINUTE);
DECLARE
  end_ts DEFAULT TIMESTAMP_ADD(CURRENT_TIMESTAMP(), INTERVAL -20 MINUTE);
WITH
  latency AS (
  SELECT
    pipeline_type,
    ingest_ts,
    publish_ts,
    TIMESTAMP_DIFF(ingest_ts, publish_ts, SECOND) latency_secs
  FROM
    pipeline_update.event
  WHERE
    publish_ts BETWEEN start_ts AND end_ts)
SELECT
  pipeline_type,
  COUNT(*) total_events,
  AVG(latency.latency_secs) average_latency_secs,
  MIN(latency.latency_secs) min_latency_secs,
  MAX(latency.latency_secs) max_latency_secs,
  STDDEV(latency.latency_secs) std_deviation
FROM
  latency
GROUP BY
  pipeline_type
ORDER BY
  pipeline_type;

Missing records

To check if there were missing records:

DECLARE
  start_ts DEFAULT TIMESTAMP_ADD(CURRENT_TIMESTAMP(), INTERVAL -40 MINUTE);
DECLARE
  end_ts DEFAULT TIMESTAMP_ADD(CURRENT_TIMESTAMP(), INTERVAL -20 MINUTE);
SELECT
  COUNT(*) missed_events,
FROM
  pipeline_update.event base
WHERE
  base.publish_ts BETWEEN start_ts AND end_ts
  AND pipeline_type = 'baseline'
  AND NOT EXISTS(
  SELECT
    *
  FROM
    pipeline_update.event main
  WHERE
    main.publish_ts BETWEEN start_ts AND end_ts
    AND pipeline_type = 'main'
    AND base.id = main.id);

Duplicates

Duplicate events

DECLARE
  start_ts DEFAULT TIMESTAMP_ADD(CURRENT_TIMESTAMP(), INTERVAL -40 MINUTE);
DECLARE
  end_ts DEFAULT TIMESTAMP_ADD(CURRENT_TIMESTAMP(), INTERVAL -20 MINUTE);

SELECT
  id,
  pipeline_type,
  COUNT(*) event_count,
FROM
  pipeline_update.event base
WHERE
  base.publish_ts BETWEEN start_ts AND end_ts
GROUP BY id, pipeline_type
HAVING event_count > 1

Duplicate statistics

DECLARE
  start_ts DEFAULT TIMESTAMP_ADD(CURRENT_TIMESTAMP(), INTERVAL -40 MINUTE);
DECLARE
  end_ts DEFAULT TIMESTAMP_ADD(CURRENT_TIMESTAMP(), INTERVAL -20 MINUTE);
WITH
  counts AS (
  SELECT
    COUNT(id) total_event_count,
    COUNT(DISTINCT id) event_distinct_count,
    pipeline_type,
  FROM
    pipeline_update.event base
  WHERE
    base.publish_ts BETWEEN start_ts
    AND end_ts
  GROUP BY
    pipeline_type)
SELECT
  event_distinct_count,
  counts.total_event_count - counts.event_distinct_count AS dups_count,
  (counts.total_event_count - counts.event_distinct_count)*100/counts.event_distinct_count dups_percentage,
  pipeline_type
FROM
  counts
ORDER BY
  pipeline_type DESC;

Cleanup

./stop-event-generation.sh
./stop-processing-pipelines.sh
terraform -chdir terraform destroy 

Contributing

Contributions to this repo are always welcome and highly encouraged.

See CONTRIBUTING for more information how to get started.

Please note that this project is released with a Contributor Code of Conduct. By participating in this project you agree to abide by its terms. See Code of Conduct for more information.

License

Apache 2.0 - See LICENSE for more information.