fabric_ceph_client

Python client for the FABRIC Ceph Manager service, which brokers access to FABRIC's distributed Ceph clusters: CephFS subvolumes and CephX users for POSIX storage, and RGW users, keys and buckets for S3.

Its only dependency is requests.

Install

pip install fabric_ceph_client

Usage

Authenticate with a FABRIC token, either inline or from a token file:

from fabric_ceph_client.fabric_ceph_client import CephManagerClient

client = CephManagerClient(
    base_url="https://ceph-mgr.fabric-testbed.net",
    token_file="~/work/fabric_config/id_token.json",
)

Every call targets a single cluster, passed as cluster. Each cluster is an independent Ceph deployment, so the same identity in two clusters is two separate accounts.

Cluster info

info = client.list_cluster_info()
for c in info["data"]:
    print(c["cluster"], c["mon_host"], c.get("s3_endpoints"))

CephFS

client.create_or_resize_subvolume(
    "east", "CEPH-FS-01", "alice", size=10 * 1024**3
)
client.get_subvolume_info("east", "CEPH-FS-01", "alice")
client.list_subvolumes("east", "CEPH-FS-01")
client.delete_subvolume("east", "CEPH-FS-01", "alice")

S3 (RGW)

An S3 user id is the user's FABRIC bastion login, the same identity used to name their CephFS subvolume.

# Users and access keys
client.list_s3_users("east")
client.get_s3_user("east", "alice_0000123456")

# Secrets are withheld from listings unless asked for explicitly
client.list_s3_user_keys("east", "alice_0000123456", include_secret=True)

# Creating a key also provisions the S3 user on first use
key = client.create_s3_user_key("east", "alice_0000123456")
print(key["access_key"], key["secret_key"])

# Buckets
client.list_s3_buckets("east", uid="alice_0000123456")
client.get_s3_bucket("east", "my-bucket")
client.set_s3_bucket_quota("east", "my-bucket", enabled=True, max_size_kb=1024**2)

Creating and deleting buckets is restricted to FABRIC facility administrators and owners of the FABRIC Ceph service project. Regular users can read and write objects in the buckets they own, using the credentials above with any S3 client (aws-cli, s3cmd, boto3, rclone). Object data goes directly to the Ceph RGW gateway and never passes through the Ceph Manager service.

Errors

Failed requests raise ApiError, which carries the HTTP status, the URL, and the parsed response body:

from fabric_ceph_client.fabric_ceph_client import ApiError

try:
    client.get_s3_bucket("east", "nope")
except ApiError as e:
    print(e.status, e.message, e.payload)

Links

• Service source: https://github.com/fabric-testbed/fabric_ceph
• FABRIC: https://fabric-testbed.net/

License

Apache-2.0