nbdkit-gcs-plugin1

Usage

Overview

"nbdkit-gcs-plugin" is a plugin for nbdkit(1) which lets you open objects stored in Google Cloud Storage (GCS) as disk images.

This plugin uses the Python Google Cloud Storage client library (google-cloud-storage). It requires Python ≥ 3.7.

Examples

 nbdkit gcs json-credentials=/path/to-my-json-credentials.json \
            bucket=MY-BUCKET key=disk.img

Provides a read only block device holding the data contained in the disk.img object.

 nbdkit gcs json-credentials=/path/to-my-json-credentials.json \
            size=50G object-size=128k \
            bucket=MY-BUCKET key=disk

Provides a read-write block device with size 50G, whose contents are stored across multiple in objects of size 128k, prefixed with disk/

Parameters

json-credentials=FILE

Path to a JSON file with Google Service account keys. Application Default Credentials will be used if "json-credentials" is not specified (see "CREDENTIALS" below).

bucket=BUCKET

The bucket containing the object(s). This parameter is required.

key=STRING

The object name (if "size" is not specified) or object prefix (if "size" is specified) to use within the bucket. This parameter is required.

size=SIZE

object-size=SIZE

These two parameters must always be specified together. If set, data will be split into blocks of "object-size" and stored as separate objects. The block device will report a total size of "size" and be writeable and trim-able.

Object names will have the form key/%16x, where %16x is the 16-digit hexadecimal block number. If there are existing objects under the prefix that do not have the expected size, the plugin will crash.

Performance considerations

It is highly recommended that clients do their utmost to issue requests that exactly match the object size: Smaller write requests will incur a performance penalty due to the need for read-modify-write cycles (thus also incurring latency from two network round-trips). Larger read and write requests will incur a performance penalty because of sequential execution.

The nbdkit-blocksize-filter(1) can be used to alleviate the impact of requests larger than the object size, but does not help if the client issues requests smaller than the block size.

The nbdkit-stats-filter(1) can be used to investigate what block sizes and alignments are used by the client.

When connecting through the Linux kernel's NBD module, consider setting "/sys/block/nbd<X>/queue/max_sectors_kb" to match the object size.

Credentials

Refer to the following web pages:

Google Service account keys

https://developers.google.com/workspace/guides/create-credentials#service-account

Application Default Credentials

https://cloud.google.com/docs/authentication/application-default-credentials

Files

$plugindir/nbdkit-gcs-plugin

The plugin.

Use "nbdkit --dump-config" to find the location of $plugindir.

$HOME/.config/gcloud/application_default_credentials.json

Application Default Credentials.

Environment

"GOOGLE_APPLICATION_CREDENTIALS"

You can use the "GOOGLE_APPLICATION_CREDENTIALS" environment variable to provide the location of a credential JSON file.

Comparison to S3 plugin

You can use the "S3" plugin to access Google Cloud Storage in interoperability mode. "read" and "write" work perfectly fine, while "trim" and "zero" operations don't work in the general case: trimmed or zeroed-out blocks cannot be deleted from the storage using the "S3" plugin.

Version

"nbdkit-gcs-plugin" first appeared in nbdkit 1.38.