• This repository has been archived on 11/Feb/2018
  • Stars
    star
    135
  • Rank 269,241 (Top 6 %)
  • Language
    Go
  • License
    MIT License
  • Created over 9 years ago
  • Updated about 7 years ago

Reviews

There are no reviews yet. Be the first to send feedback to the community and the maintainers!

Repository Details

Stream data into Google BigQuery concurrently using InsertAll()

Kik and me (@oryband) are no longer maintaining this repository. Thanks for all the contributions. You are welcome to fork and continue development.

BigQuery Streamer BigQuery GoDoc

Stream insert data into BigQuery fast and concurrently, using InsertAll().

Features

  • Insert rows from multiple tables, datasets, and projects, and insert them bulk. No need to manage data structures and sort rows by tables - bqstreamer does it for you.
  • Multiple background workers (i.e. goroutines) to enqueue and insert rows.
  • Insert can be done in a blocking or in the background (asynchronously).
  • Perform insert operations in predefined set sizes, according to BigQuery's quota policy.
  • Handle and retry BigQuery server errors.
  • Backoff interval between failed insert operations.
  • Error reporting.
  • Production ready, and thoroughly tested. We - at Rounds (now acquired by Kik) - are using it in our data gathering workflow.
  • Thorough testing and documentation for great good!

Getting Started

  1. Install Go, version should be at least 1.5.
  2. Clone this repository and download dependencies:
  3. Version v2: go get gopkg.in/kikinteractive/go-bqstreamer.v2
  4. Version v1: go get gopkg.in/kikinteractive/go-bqstreamer.v1
  5. Acquire Google OAuth2/JWT credentials, so you can authenticate with BigQuery.

How Does It Work?

There are two types of inserters you can use:

  1. SyncWorker, which is a single blocking (synchronous) worker.
  2. It enqueues rows and performs insert operations in a blocking manner.
  3. AsyncWorkerGroup, which employes multiple background SyncWorkers.
  4. The AsyncWorkerGroup enqueues rows, and its background workers pull and insert in a fan-out model.
  5. An insert operation is executed according to row amount or time thresholds for each background worker.
  6. Errors are reported to an error channel for processing by the user.
  7. This provides a higher insert throughput for larger scale scenarios.

Examples

Check the GoDoc examples section.

Contribute

  1. Please check the issues page.
  2. File new bugs and ask for improvements.
  3. Pull requests welcome!

Test

# Run unit tests and check coverage.
$ make test

# Run integration tests.
# This requires an active project, dataset and pem key.
$ export BQSTREAMER_PROJECT=my-project
$ export BQSTREAMER_DATASET=my-dataset
$ export BQSTREAMER_TABLE=my-table
$ export BQSTREAMER_KEY=my-key.json
$ make testintegration