Skip to content

A Golang application to export table level metrics from BigQuery into Datadog.

License

Notifications You must be signed in to change notification settings

ovotech/bigquery-metrics-exporter

Repository files navigation

bigquery-metrics-exporter

A Golang application to export table level metrics from BigQuery into Datadog.

Two binaries are provided. bqmetrics runs a single round of metrics collection. bqmetricsd runs metrics collection continually according to the provided metric collection interval.

Metrics

The metrics exporter queries the BigQuery metadata API to generate metrics. The metadata API only stores this information for tables and materialized views, so views and external data sources will not have metrics exported.

The following metrics are generated:

  • row_count - The number of rows in the table
  • last_modified - The number of seconds since this table was last modified
  • last_modified_time - The timestamp when the table was last modified

Inserting or modifying data in the table also updates the last modified time, so those metrics can be used as a measure of data freshness.

Custom Metrics

The metrics exporter also includes the ability to generate Datadog metrics from the results of SQL queries.

⚠️ Running an SQL query on BigQuery may have a cost associated with it

Each custom metric has a name, a list of tags, and its own collection interval as well as the SQL query to run to produce the metrics. The SQL query should return a single row of data, and each column will be exported as a distinct metric.

Recommended usage

It is recommended to run the metrics collection daemon bqmetricsd which will continually collect metrics and ship them to Datadog according to the provided schedule.

bqmetricsd \
  --datadog-api-key-file datadog.key \
  --gcp-project-id my-project \
  --metric-interval 1m \
  --metric-tags team:myteam,env:prod

Running in Google Cloud Platform

Running in Google Cloud Platform is the preferred method of operation as it will reduce latency for metrics collection and simplify authentication to the BigQuery API. A Terraform module is provided to simplify running the daemon in GCP. Example usage is below:

data "google_compute_subnetwork" "default" {
  name   = "default"
  region = "europe-west1"
}

module "bqmetrics" {
  source = "git::https://github.com/ovotech/bigquery-metrics-exporter.git//terraform/gcp?ref=v1.2.2"

  datadog-api-key-secret = "datadog-api-key"
  subnetwork             = data.google_compute_subnetwork.default.self_link
}

The Terraform provider makes use of Google Secrets Manager to handle the Datadog API secret key. This secret can be created with the gcloud CLI utility using the following command:

printf "secret" | gcloud secrets create datadog-api-key --data-file=-

Depending on organizational policy, you may need to restrict the secret to certain locations. See gcloud secrets create --help for full details.

An existing secret can be updated with the following commands:

printf "secret" | gcloud secrets versions add datadog-api-key --data-file=-

Configuration

bqmetrics and bqmetricsd are both configurable using the same mechanisms, either a config file, environment variables, or on the command line. Config set on the command line has priority over environment variables, which in turn have priority over the config file.

It is required that the Datadog API key is set using one of the available options in order to run. Credentials also need to be provided for connecting to the GCP APIs, although that may be handled automatically by the environment. See the Google Cloud Platform authentication documentation for more information. The Google Cloud Project ID is also required. All other parameters are optional.

Config file

See example-config.yaml for an example config file that details all the current configuration.

bqmetrics and bqmetricsd will by default search for a config file at /etc/bqmetrics/config.yaml and ~/.bqmetrics/config.yaml, although you can also specify the path to a config file using the --config-file command line parameter or the CONFIG_FILE environment variable.

Environment and command line parameters

Below is a list of configuration available as environment variables and command line options.

Environment Variable Parameter Description
CONFIG_FILE --config-file Path to the config file
DATADOG_API_KEY The Datadog API key
DATADOG_API_KEY_FILE --datadog-api-key-file File containing Datadog API key
DATADOG_API_KEY_SECRET_ID --datadog-api-key-secret-id Path to a secret held in Google Secret Manager containing Datadog API key, e.g. projects/my-project/secrets/datadog-api-key/versions/3
DATASET_FILTER --dataset-filter BigQuery label to filter datasets for metric collection
GCP_PROJECT_ID --gcp-project-id (Required) The Google Cloud project containing the BigQuery tables to retrieve metrics from
GOOGLE_APPLICATION_CREDENTIALS File containing service account details to authenticate to Google Cloud using
HEALTHCHECK_ENABLED --healthcheck.enabled Whether to enable the health check endpoint at /health. Defaults to false
HEALTHCHECK_PORT --healthcheck.port The port to run the health check server on. Defaults to 8080
LOG_LEVEL The logging level (e.g. trace, debug, info, warn, error). Defaults to info
METRIC_INTERVAL --metric-interval The interval between metric collection rounds. Must contain a unit and valid units are "ns", "us" (or "µs"), "ms", "s", "m", "h". Defaults to 30s
METRIC_PREFIX --metric-prefix The prefix for the metric names exported to Datadog. Defaults to custom.gcp.bigquery
METRIC_TAGS --metric-tags Comma-delimited list of tags to attach to metrics (e.g. env:prod,team:myteam)

GCP Service Account permissions

The service account running bqmetricsd may require the following roles:

BigQuery Data Viewer
    Required to generate custom metrics that need access to table data
    This permission can be granted directly on the datasets in question 
BigQuery Metadata Viewer
    Required to generate table level metrics
BigQuery User
    Required to generate custom metrics
Secret Manager Secret Accessor
    Required to access the Datadog API key if stored in Secret Manager
    This permission can be granted directly on the secret in question

About

A Golang application to export table level metrics from BigQuery into Datadog.

Topics

Resources

License

Stars

Watchers

Forks

Packages

No packages published

Contributors 3

  •  
  •  
  •  

Languages