cloudevent-recorder/

README

cloudevent-recorder

This module provisions a regionalized event subscriber that consumes events of particular types from each of the regional brokers, writes them to Google BigQuery.

flowchart LR
    subgraph "regional network"
    A[[Pub/Sub topic]] -- notifies --> B(Recorder Service)
    end

   B -- writes every 3m --> C[(Cloud Storage)]
   C --> D[Data Transfer Service]
   D -- loads every 15m --> E[(BigQuery)]

The recorder service writes to a regional GCS bucket where a periodic BigQuery Data-Transfer Service Job will pull events from into a BigQuery table schematized for that event type.

The intended usage of this module for publishing events is something like this:

// Create a network with several regional subnets
module "networking" {
  source = "chainguard-dev/common/infra//modules/networking"

  name       = "my-networking"
  project_id = var.project_id
  regions    = [...]
}

// Create the Broker abstraction.
module "cloudevent-broker" {
  source = "chainguard-dev/common/infra//modules/cloudevent-broker"

  name       = "my-broker"
  project_id = var.project_id
  regions    = module.networking.regional-networks
}

// Record cloudevents of type com.example.foo and com.example.bar
module "foo-emits-events" {
  source = "chainguard-dev/common/infra//modules/cloudevent-recorder"

  name       = "my-recorder"
  project_id = var.project_id
  regions    = module.networking.regional-networks
  broker     = module.cloudevent-broker.broker

  retention-period = 30 // keep around 30 days worth of event data

  provisioner = "user:sally@chainguard.dev"

  types = {
    "com.example.foo": {
      schema = file("${path.module}/foo.schema.json")
    }
    "com.example.bar": {
      schema = file("${path.module}/bar.schema.json")
    }
  }
}

The default behavior of this module is to deploy a cloud event trigger that consumes events from a broker and uses log rotate to write them to a GCS bucket. The GCS bucket is then used as a source for a BigQuery Data Transfer Service job.

To override this behavior, you can choose a different method.

i.e. to use GCP native integration for pubsub to GCS:

module "foo-emits-events" {
  source = "chainguard-dev/common/infra//modules/cloudevent-recorder"
  method = "gcs"

  ...
}

Requirements

No requirements.

Providers

Name	Version
google	n/a
google-beta	n/a
random	n/a

Modules

Name	Source	Version
audit-import-serviceaccount	../audit-serviceaccount	n/a
recorder-dashboard	../dashboard/cloudevent-receiver	n/a
this	../regional-go-service	n/a
triggers	../cloudevent-trigger	n/a

Resources

Name	Type
google-beta_google_project_service_identity.pubsub	resource
google_bigquery_data_transfer_config.import-job	resource
google_bigquery_dataset.this	resource
google_bigquery_table.types	resource
google_bigquery_table_iam_binding.import-writes-to-tables	resource
google_monitoring_alert_policy.bq_dts	resource
google_pubsub_subscription.dead-letter-pull-sub	resource
google_pubsub_subscription.this	resource
google_pubsub_subscription_iam_binding.allow-pubsub-to-ack	resource
google_pubsub_topic.dead-letter	resource
google_pubsub_topic_iam_binding.allow-pubsub-to-send-to-dead-letter	resource
google_service_account.import-identity	resource
google_service_account.recorder	resource
google_service_account_iam_binding.bq-dts-assumes-import-identity	resource
google_service_account_iam_binding.provisioner-acts-as-import-identity	resource
google_storage_bucket.recorder	resource
google_storage_bucket_iam_binding.broker-writes-to-gcs-buckets	resource
google_storage_bucket_iam_binding.import-reads-from-gcs-buckets	resource
google_storage_bucket_iam_binding.recorder-writes-to-gcs-buckets	resource
random_id.suffix	resource
random_id.trigger-suffix	resource
random_string.suffix	resource
google_client_openid_userinfo.me	data source
google_project.project	data source

Inputs

Name	Description	Type	Default	Required
ack_deadline_seconds	The number of seconds to acknowledge a message before it is redelivered.	number	300	no
broker	A map from each of the input region names to the name of the Broker topic in that region.	map(string)	n/a	yes
cloud_storage_config_max_bytes	The maximum bytes that can be written to a Cloud Storage file before a new file is created. Min 1 KB, max 10 GiB.	number	1000000000	no
cloud_storage_config_max_duration	The maximum duration that can elapse before a new Cloud Storage file is created. Min 1 minute, max 10 minutes, default 5 minutes.	number	300	no
deletion_protection	Whether to enable deletion protection on data resources.	bool	true	no
enable_profiler	Enable cloud profiler.	bool	false	no
flush_interval	Flush interval for logrotate, as a duration string.	string	""	no
ignore_unknown_values	Whether to ignore unknown values in the data, when transferring data to BigQuery.	bool	false	no
limits	Resource limits for the regional go service.	object({ cpu = string memory = string })	null	no
location	The location to create the BigQuery dataset in, and in which to run the data transfer jobs from GCS.	string	"US"	no
max_delivery_attempts	The maximum number of delivery attempts for any event.	number	5	no
maximum_backoff	The maximum delay between consecutive deliveries of a given message.	number	600	no
method	The method used to transfer events (e.g., trigger, gcs).	string	"trigger"	no
minimum_backoff	The minimum delay between consecutive deliveries of a given message.	number	10	no
name	n/a	string	n/a	yes
notification_channels	List of notification channels to alert (for service-level issues).	list(string)	n/a	yes
project_id	n/a	string	n/a	yes
provisioner	The identity as which this module will be applied (so it may be granted permission to 'act as' the DTS service account). This should be in the form expected by an IAM subject (e.g. user:sally@example.com)	string	n/a	yes
regions	A map from region names to a network and subnetwork. A recorder service and cloud storage bucket (into which the service writes events) will be created in each region.	map(object({ network = string subnet = string }))	n/a	yes
retention-period	The number of days to retain data in BigQuery.	number	n/a	yes
scaling	The scaling configuration for the service.	object({ min_instances = optional(number, 0) max_instances = optional(number, 100) max_instance_request_concurrency = optional(number) })	{}	no
split_triggers	Opt-in flag to split into per-trigger dashboards. Helpful when hitting widget limits	bool	false	no
types	A map from cloudevent types to the BigQuery schema associated with them, as well as an alert threshold and a list of notification channels (for subscription-level issues).	map(object({ schema = string alert_threshold = optional(number, 50000) notification_channels = optional(list(string), []) partition_field = optional(string) }))	n/a	yes

Outputs

Name	Description
dataset_id	n/a
table_ids	n/a