Use GCS offloader with Pulsar

This chapter guides you through every step of installing and configuring the GCS offloader and using it with Pulsar.

Installation

Follow the steps below to install the GCS offloader.

Prerequisite

  • Pulsar: 2.4.2 or later versions

Step

This example uses Pulsar 2.5.1.

  1. Download the Pulsar tarball using one of the following ways:

    • Download from the Apache mirror

    • Download from the Pulsar download page

    • Use wget

      1. wget https://archive.apache.org/dist/pulsar/pulsar-2.5.1/apache-pulsar-2.5.1-bin.tar.gz
  1. Download and untar the Pulsar offloaders package.

    1. wget https://downloads.apache.org/pulsar/pulsar-2.5.1/apache-pulsar-offloaders-2.5.1-bin.tar.gz
    2. tar xvfz apache-pulsar-offloaders-2.5.1-bin.tar.gz
  1. ##### note
  2. - If you are running Pulsar in a bare metal cluster, make sure that `offloaders` tarball is unzipped in every broker's Pulsar directory.
  3. - If you are running Pulsar in Docker or deploying Pulsar using a Docker image (such as K8S and DCOS), you can use the `apachepulsar/pulsar-all` image instead of the `apachepulsar/pulsar` image. `apachepulsar/pulsar-all` image has already bundled tiered storage offloaders.
  1. Copy the Pulsar offloaders as offloaders in the Pulsar directory.

    1. mv apache-pulsar-offloaders-2.5.1/offloaders apache-pulsar-2.5.1/offloaders
    2. ls offloaders
  1. **Output**
  2. As shown in the output, Pulsar uses [Apache jclouds](https://jclouds.apache.org) to support GCS and AWS S3 for long term storage.
  3. ```
  4. tiered-storage-file-system-2.5.1.nar
  5. tiered-storage-jcloud-2.5.1.nar
  6. ```

Configuration

note

Before offloading data from BookKeeper to GCS, you need to configure some properties of the GCS offloader driver.

Besides, you can also configure the GCS offloader to run it automatically or trigger it manually.

Configure GCS offloader driver

You can configure GCS offloader driver in the configuration file broker.conf or standalone.conf.

  • Required configurations are as below.

    Required configurationDescriptionExample value
    managedLedgerOffloadDriverOffloader driver name, which is case-insensitive.google-cloud-storage
    offloadersDirectoryOffloader directoryoffloaders
    gcsManagedLedgerOffloadBucketBucketpulsar-topic-offload
    gcsManagedLedgerOffloadRegionBucket regioneurope-west3
    gcsManagedLedgerOffloadServiceAccountKeyFileAuthentication/Users/user-name/Downloads/project-804d5e6a6f33.json
  • Optional configurations are as below.

    Optional configurationDescriptionExample value
    gcsManagedLedgerOffloadReadBufferSizeInBytesSize of block read1 MB
    gcsManagedLedgerOffloadMaxBlockSizeInBytesSize of block write64 MB
    managedLedgerMinLedgerRolloverTimeMinutesMinimum time between ledger rollover for a topic.2
    managedLedgerMaxEntriesPerLedgerThe max number of entries to append to a ledger before triggering a rollover.5000

Bucket (required)

A bucket is a basic container that holds your data. Everything you store in GCS must be contained in a bucket. You can use a bucket to organize your data and control access to your data, but unlike directory and folder, you can not nest a bucket.

Example

This example names the bucket as pulsar-topic-offload.

  1. gcsManagedLedgerOffloadBucket=pulsar-topic-offload

Bucket region (required)

Bucket region is the region where a bucket is located. If a bucket region is not specified, the default region (us multi-regional location) is used.

tip

For more information about bucket location, see here.

Example

This example sets the bucket region as europe-west3.

  1. gcsManagedLedgerOffloadRegion=europe-west3

Authentication (required)

To enable a broker access GCS, you need to configure gcsManagedLedgerOffloadServiceAccountKeyFile in the configuration file broker.conf.

gcsManagedLedgerOffloadServiceAccountKeyFile is a JSON file, containing GCS credentials of a service account.

Example

To generate service account credentials or view the public credentials that you’ve already generated, follow the following steps.

  1. Navigate to the Service accounts page.

  2. Select a project or create a new one.

  3. Click Create service account.

  4. In the Create service account window, type a name for the service account and select Furnish a new private key.

    If you want to grant G Suite domain-wide authority to the service account, select Enable G Suite Domain-wide Delegation.

  5. Click Create.

    note

    Make sure the service account you create has permission to operate GCS, you need to assign Storage Admin permission to your service account here.

  6. You can get the following information and set this in broker.conf.

    1. gcsManagedLedgerOffloadServiceAccountKeyFile="/Users/user-name/Downloads/project-804d5e6a6f33.json"
  1. ##### tip
  2. - For more information about how to create `gcsManagedLedgerOffloadServiceAccountKeyFile`, see [here](https://support.google.com/googleapi/answer/6158849).
  3. - For more information about Google Cloud IAM, see [here](https://cloud.google.com/storage/docs/access-control/iam).

Size of block read/write

You can configure the size of a request sent to or read from GCS in the configuration file broker.conf.

ConfigurationDescription
gcsManagedLedgerOffloadReadBufferSizeInBytesBlock size for each individual read when reading back data from GCS.

The default value is 1 MB.
gcsManagedLedgerOffloadMaxBlockSizeInBytesMaximum size of a “part” sent during a multipart upload to GCS.

It can not be smaller than 5 MB.

The default value is 64 MB.

Configure GCS offloader to run automatically

Namespace policy can be configured to offload data automatically once a threshold is reached. The threshold is based on the size of data that a topic has stored on a Pulsar cluster. Once the topic reaches the threshold, an offload operation is triggered automatically.

Threshold valueAction
> 0It triggers the offloading operation if the topic storage reaches its threshold.
= 0It causes a broker to offload data as soon as possible.
< 0It disables automatic offloading operation.

Automatic offloading runs when a new segment is added to a topic log. If you set the threshold on a namespace, but few messages are being produced to the topic, offloader does not work until the current segment is full.

You can configure the threshold size using CLI tools, such as pulsar-admin.

The offload configurations in broker.conf and standalone.conf are used for the namespaces that do not have namespace level offload policies. Each namespace can have its own offload policy. If you want to set offload policy for each namespace, use the command pulsar-admin namespaces set-offload-policies options command.

Example

This example sets the GCS offloader threshold size to 10 MB using pulsar-admin.

  1. pulsar-admin namespaces set-offload-threshold --size 10M my-tenant/my-namespace
tip

For more information about the pulsar-admin namespaces set-offload-threshold options command, including flags, descriptions, default values, and shorthands, see here.

Configure GCS offloader to run manually

For individual topics, you can trigger GCS offloader manually using one of the following methods:

  • Use REST endpoint.

  • Use CLI tools (such as pulsar-admin).

    To trigger the GCS via CLI tools, you need to specify the maximum amount of data (threshold) that should be retained on a Pulsar cluster for a topic. If the size of the topic data on the Pulsar cluster exceeds this threshold, segments from the topic are moved to GCS until the threshold is no longer exceeded. Older segments are moved first.

Example

  • This example triggers the GCS offloader to run manually using pulsar-admin with the command pulsar-admin topics offload (topic-name) (threshold).

    1. pulsar-admin topics offload persistent://my-tenant/my-namespace/topic1 10M
  1. **Output**
  2. ```
  3. Offload triggered for persistent://my-tenant/my-namespace/topic1 for messages before 2:0:-1
  4. ```
  5. ##### tip
  6. For more information about the `pulsar-admin topics offload options` command, including flags, descriptions, default values, and shorthands, see [here](https://pulsar.apache.org/docs/pulsar-admin#offload).
  • This example checks the GCS offloader status using pulsar-admin with the command pulsar-admin topics offload-status options.

    1. pulsar-admin topics offload-status persistent://my-tenant/my-namespace/topic1
  1. **Output**
  2. ```
  3. Offload is currently running
  4. ```
  5. To wait for GCS to complete the job, add the `-w` flag.
  6. ```
  7. pulsar-admin topics offload-status -w persistent://my-tenant/my-namespace/topic1
  8. ```
  9. **Output**
  10. ```
  11. Offload was a success
  12. ```
  13. If there is an error in offloading, the error is propagated to the `pulsar-admin topics offload-status` command.
  14. ```
  15. pulsar-admin topics offload-status persistent://my-tenant/my-namespace/topic1
  16. ```
  17. **Output**
  18. ```
  19. Error in offload
  20. null
  21. Reason: Error offloading: org.apache.bookkeeper.mledger.ManagedLedgerException: java.util.concurrent.CompletionException: com.amazonaws.services.s3.model.AmazonS3Exception: Anonymous users cannot initiate multipart uploads. Please authenticate. (Service: Amazon S3; Status Code: 403; Error Code: AccessDenied; Request ID: 798758DE3F1776DF; S3 Extended Request ID: dhBFz/lZm1oiG/oBEepeNlhrtsDlzoOhocuYMpKihQGXe6EG8puRGOkK6UwqzVrMXTWBxxHcS+g=), S3 Extended Request ID: dhBFz/lZm1oiG/oBEepeNlhrtsDlzoOhocuYMpKihQGXe6EG8puRGOkK6UwqzVrMXTWBxxHcS+g=
  22. ```
  23. ##### tip
  24. For more information about the `pulsar-admin topics offload-status options` command, including flags, descriptions, default values, and shorthands, see [here](https://pulsar.apache.org/docs/pulsar-admin#offload-status).

Tutorial

For the complete and step-by-step instructions on how to use the GCS offloader with Pulsar, see here.