Google Cloud Storage Input

Use the google cloud storage input to read content from files stored in buckets which reside on your Google Cloud. The input can be configured to work with and without polling, though currently, if polling is disabled it will only perform a one time passthrough, list the file contents and end the process. Polling is generally recommented for most cases even though it can get expensive with dealing with a very large number of files.

To mitigate errors and ensure a stable processing environment, this input employs the following features :

A sample configuration with detailed explanation for each field is given below :-

filebeat.inputs:
- type: gcs
  id: my-gcs-id
  enabled: true
  project_id: my_project_id
  auth.credentials_file.path: {{file_path}}/{{creds_file_name}}.json
  parse_json: true
  buckets:
  - name: gcs-test-new
    max_workers: 3
    poll: true
    poll_interval: 15s
    bucket_timeout: 60s
  - name: gcs-test-old
    max_workers: 3
    poll: true
    poll_interval: 10s
    bucket_timeout: 30s

Explanation : This configuration given above describes a basic gcs config having two buckets named gcs-test-new and gcs-test-old. Each of these buckets have their own attributes such as name, max_workers, poll, poll_interval and bucket_timeout. These attributes have detailed explanations given below. For now lets try to understand how this config works.

For google cloud storage input to identify the files it needs to read and process, it will require the bucket names to be specified. We can have as many buckets as we deem fit. We are also able to configure the attributes max_workers, poll, poll_interval and bucket_timeout at the root level, which will then be applied to all buckets which do not specify any of these attributes explicitly.

On receiving this config the google cloud storage input will connect to the service and retrieve a Storage Client using the given bucket_name and auth.credentials_file, then it will spawn two main go-routines, one for each bucket. After this each of these routines (threads) will initialize a scheduler which will in turn use the max_workers value to initialize an in-memory worker pool (thread pool) with 3 workers available. Basically that equates to two instances of a worker pool, one per bucket, each having 3 workers. These workers will be responsible for performing jobs that process a file (in this case read and output the contents of a file).

A Sample Response :-

{
  "@timestamp": "2022-09-01T13:54:24.588Z",
  "@metadata": {
    "beat": "filebeat",
    "type": "_doc",
    "version": "8.5.0",
    "_id": "gcs-test-new-data_3.json-worker-1"
  },
  "log": {
    "offset": 200,
    "file": {
      "path": "gs://gcs-test-new/data_3.json"
    }
  },
  "input": {
    "type": "gcs"
  },
  "message": "{\n    \"id\": 1,\n    \"title\": \"iPhone 9\",\n    \"description\": \"An apple mobile which is nothing like apple\",\n    \"price\": 549,\n    \"discountPercentage\": 12.96,\n    \"rating\": 4.69,\n    \"stock\": 94,\n    \"brand\": \"Apple\",\n    \"category\": \"smartphones\",\n    \"thumbnail\": \"https://dummyjson.com/image/i/products/1/thumbnail.jpg\",\n    \"images\": [\n        \"https://dummyjson.com/image/i/products/1/1.jpg\",\n        \"https://dummyjson.com/image/i/products/1/2.jpg\",\n        \"https://dummyjson.com/image/i/products/1/3.jpg\",\n        \"https://dummyjson.com/image/i/products/1/4.jpg\",\n        \"https://dummyjson.com/image/i/products/1/thumbnail.jpg\"\n    ]\n}\n",
  "cloud": {
    "provider": "goole cloud"
  },
  "gcs": {
    "storage": {
      "bucket": {
        "name": "gcs-test-new"
      },
      "object": {
        "name": "data_3.json",
        "content_type": "application/json",
        "json_data": [
          {
            "id": 1,
            "discountPercentage": 12.96,
            "rating": 4.69,
            "brand": "Apple",
            "price": 549,
            "category": "smartphones",
            "thumbnail": "https://dummyjson.com/image/i/products/1/thumbnail.jpg",
            "description": "An apple mobile which is nothing like apple",
            "title": "iPhone 9",
            "stock": 94,
            "images": [
              "https://dummyjson.com/image/i/products/1/1.jpg",
              "https://dummyjson.com/image/i/products/1/2.jpg",
              "https://dummyjson.com/image/i/products/1/3.jpg",
              "https://dummyjson.com/image/i/products/1/4.jpg",
              "https://dummyjson.com/image/i/products/1/thumbnail.jpg"
            ]
          }
        ]
      }
    }
  },
  "event": {
    "kind": "publish_data"
  }
}

As we can see from the response above, the message field contains the original stringified data while the gcs.storage.object.data contains the objectified data.

Some of the key attributes are as follows :-

Now let’s explore the configuration attributes a bit more elaborately.

Supported Attributes :-

This attribute is required for various internal operations with respect to authentication, creating storage clients and logging which are used internally for various processing purposes.

This attribute contains the json service account credentials string, which can be generated from the google cloud console, ref: https://cloud.google.com/iam/docs/creating-managing-service-account-keys, under the respective storage account. A single storage account can contain multiple buckets, and they will all use this common service account access key.

This attribute contains the service account credentials file, which can be generated from the google cloud console, ref: https://cloud.google.com/iam/docs/creating-managing-service-account-keys, under the respective storage account. A single storage account can contain multiple buckets, and they will all use this common service account credentials file.

This attribute contains the details about a specific bucket like name, max_workers, poll, poll_interval and bucket_timeout. The attribute name is specific to a bucket as it describes the bucket name, while the fields max_workers, poll, poll_interval and bucket_timeout can exist both at the bucket level and the root level. This attribute is internally represented as an array, so we can add as many buckets as we require.

This is a specific subfield of a bucket. It specifies the bucket name.

This attribute defines the maximum amount of time after which a bucket operation will give and stop if no response is recieved (example: reading a file / listing a file). It can be defined in the following formats : {{x}}s, {{x}}m, {{x}}h, here s = seconds, m = minutes and h = hours. The value {{x}} can be anything we wish. If no value is specified for this, by default its initialized to 50 seconds. This attribute can be specified both at the root level of the configuration as well at the bucket level. The bucket level values will always take priority and override the root level values if both are specified. The value of bucket_timeout that should be used depends on the size of the files and the network speed. If the timeout is too low, the input will not be able to read the file completely and context_deadline_exceeded errors will be seen in the logs. If the timeout is too high, the input will wait for a long time for the file to be read, which can cause the input to be slow. The ratio between the bucket_timeout and poll_interval should be considered while setting both the values. A low poll_interval and a very high bucket_timeout can cause resource utilization issues as schedule ops will be spawned every poll iteration. If previous poll ops are still running, this could result in concurrently running ops and so could cause a bottleneck over time.

This attribute defines the maximum number of workers (goroutines / lightweight threads) are allocated in the worker pool (thread pool) for processing jobs which read the contents of files. This attribute can be specified both at the root level of the configuration and at the bucket level. Bucket level values override the root level values if both are specified. Larger number of workers do not necessarily improve of throughput, and this should be carefully tuned based on the number of files, the size of the files being processed and resources available. Increasing max_workers to very high values may cause resource utilization problems and can lead to a bottleneck in processing. Usually a maximum cap of 2000 workers is recommended. A very low max_worker count will drastically increase the number of network calls required to fetch the objects, which can cause a bottleneck in processing.

This attribute informs the scheduler whether to keep polling for new files or not. Default value of this is false, so it will not keep polling if not explicitly specified. This attribute can be specified both at the root level of the configuration as well at the bucket level. The bucket level values will always take priority and override the root level values if both are specified.

This attribute defines the maximum amount of time after which the internal scheduler will make the polling call for the next set of objects/files. It can be defined in the following formats : {{x}}s, {{x}}m, {{x}}h, here s = seconds, m = minutes and h = hours. The value {{x}} can be anything we wish. Example : 10s would mean we would like the polling to occur every 10 seconds. If no value is specified for this, by default its initialized to 300 seconds. This attribute can be specified both at the root level of the configuration as well at the bucket level. The bucket level values will always take priority and override the root level values if both are specified. The poll_interval should be set to a value that is equal to the bucket_timeout value. This would ensure that another schedule operation is not started before the current buckets have all been processed. If the poll_interval is set to a value that is less than the bucket_timeout, then the input will start another schedule operation before the current one has finished, which can cause a bottleneck over time. Having a lower poll_interval can make the input faster at the cost of more resource utilization.

This attribute informs the publisher whether to parse & objectify json data or not. By default this is set to false, since it can get expensive dealing with highly nested json data. If this is set to false the gcs.storage.object.json_data field in the response will have an empty array. This attribute is only applicable for json objects and has no effect on other types of objects. This attribute can be specified both at the root level of the configuration as well at the bucket level. The bucket level values will always take priority and override the root level values if both are specified.

The file encoding to use for reading data that contains international characters. This only applies to non-JSON logs. See encoding.

The file decoding option is used to specify a codec that will be used to decode the file contents. This can apply to any file stream data. An example config is shown below:

Currently supported codecs are given below:-

The CSV codec is used to decode RFC 4180 CSV data streams. Enabling the codec without other options will use the default codec options.

  decoding.codec.csv.enabled: true

The CSV codec supports five sub attributes to control aspects of CSV decoding. The comma attribute specifies the field separator character used by the CSV format. If it is not specified, the comma character , is used. The comment attribute specifies the character that should be interpreted as a comment mark. If it is specified, lines starting with the character will be ignored. Both comma and comment must be single characters. The lazy_quotes attribute controls how quoting in fields is handled. If lazy_quotes is true, a quote may appear in an unquoted field and a non-doubled quote may appear in a quoted field. The trim_leading_space attribute specifies that leading white space should be ignored, even if the comma character is white space. For complete details of the preceding configuration attribute behaviors, see the CSV decoder documentation The fields_names attribute can be used to specify the column names for the data. If it is absent, the field names are obtained from the first non-comment line of data. The number of fields must match the number of field names.

An example config is shown below:

  decoding.codec.csv.enabled: true
  decoding.codec.csv.comma: "\t"
  decoding.codec.csv.comment: "#"

If the GCS buckets have objects that correspond to files that Filebeat shouldn’t process, file_selectors can be used to limit the files that are downloaded. This is a list of selectors which are based on a regular expression pattern. The regular expression should match the object name or should be a part of the object name (ideally a prefix). The regular expression syntax used is [RE2](https://github.com/google/re2/wiki/Syntax). Files that don’t match any configured expression won’t be processed.This attribute can be specified both at the root level of the configuration as well at the container level. The container level values will always take priority and override the root level values if both are specified.

filebeat.inputs:
- type: gcs
  project_id: my_project_id
  auth.credentials_file.path: {{file_path}}/{{creds_file_name}}.json
  buckets:
  - name: obs-bucket
    max_workers: 3
    poll: true
    poll_interval: 15s
    bucket_timeout: 60s
    file_selectors:
    - regex: '/Monitoring/'
    - regex: 'docs/'
    - regex: '/Security-Logs/'

The file_selectors operation is performed within the agent locally, hence using this option will cause the agent to download all the files and then filter them. This can cause a bottleneck in processing if the number of files is very high. It is recommended to use this attribute only when the number of files is limited or ample resources are available.

If the file-set using this input expects to receive multiple messages bundled under a specific field or an array of objects then the config option for expand_event_list_from_field can be specified. This setting will be able to split the messages under the group value into separate events. For example, if you have logs that are in JSON format and events are found under the JSON object "Records". To split the events into separate events, the config option expand_event_list_from_field can be set to "Records". This attribute can be specified both at the root level of the configuration as well at the container level. The container level values will always take priority and override the root level values if both are specified.

{
    "Records": [
        {
            "eventVersion": "1.07",
            "eventTime": "2019-11-14T00:51:00Z",
            "region": "us-east-1",
            "eventID": "EXAMPLE8-9621-4d00-b913-beca2EXAMPLE",
        },
        {
            "eventVersion": "1.07",
            "eventTime": "2019-11-14T00:52:00Z",
            "region": "us-east-1",
            "eventID": "EXAMPLEc-28be-486c-8928-49ce6EXAMPLE",
        }
    ]
}

filebeat.inputs:
- type: gcs
  project_id: my_project_id
  auth.credentials_file.path: {{file_path}}/{{creds_file_name}}.json
  buckets:
  - name: obs-bucket
    max_workers: 3
    poll: true
    poll_interval: 15s
    bucket_timeout: 60s
    expand_event_list_from_field: Records

This attribute can be used to filter out files and objects that have a timestamp older than the specified value. The value of this attribute should be in unix epoch (seconds) format. The timestamp value is compared with the object.Updated field obtained from the object metadata. This attribute can be specified both at the root level of the configuration as well at the container level. The container level values will always take priority and override the root level values if both are specified.

filebeat.inputs:
- type: gcs
  project_id: my_project_id
  auth.credentials_file.path: {{file_path}}/{{creds_file_name}}.json
  buckets:
  - name: obs-bucket
    max_workers: 3
    poll: true
    poll_interval: 15s
    bucket_timeout: 60s
    timestamp_epoch: 1630444800

The GCS APIs don’t provide a direct way to filter files based on the timestamp, so the input will download all the files and then filter them based on the timestamp. This can cause a bottleneck in processing if the number of files are very high. It is recommended to use this attribute only when the number of files are limited or ample resources are available. This option scales vertically and not horizontally.

The sample configs below will explain the bucket level overriding of attributes a bit further :-

CASE - 1 :

Here bucket_1 is using root level attributes while bucket_2 overrides the values :

filebeat.inputs:
- type: gcs
  id: my-gcs-id
  enabled: true
  project_id: my_project_id
  auth.credentials_file.path: {{file_path}}/{{creds_file_name}}.json
  max_workers: 10
  poll: true
  poll_interval: 15s
  buckets:
  - name: bucket_1
  - name: bucket_2
    max_workers: 3
    poll: true
    poll_interval: 10s

Explanation : In this configuration bucket_1 has no sub attributes in max_workers, poll and poll_interval defined. It inherits the values for these fileds from the root level, which is max_workers = 10, poll = true and poll_interval = 15 seconds. However bucket_2 has these fields defined and it will use those values instead of using the root values.

CASE - 2 :

Here both bucket_1 and bucket_2 overrides the root values :

filebeat.inputs:
  - type: gcs
    id: my-gcs-id
    enabled: true
    project_id: my_project_id
    auth.credentials_file.path: {{file_path}}/{{creds_file_name}}.json
    max_workers: 10
    poll: true
    poll_interval: 15s
    buckets:
    - name: bucket_1
      max_workers: 5
      poll: true
      poll_interval: 10s
    - name: bucket_2
      max_workers: 5
      poll: true
      poll_interval: 10s

Explanation : In this configuration even though we have specified max_workers = 10, poll = true and poll_interval = 15s at the root level, both the buckets will override these values with their own respective values which are defined as part of their sub attibutes.