Storage Classes in Ceph S3

❓ What?

Storage classes in Ceph correspond to a certain data pool. They are a subsection of Placement Rules in Ceph S3 that link them to data pools. A storage class is unique to a placement rule but placement rules can have many storage classes. Every placement rule has a storage class called STANDARD by default but this STANDARD might correspond to different data pools for different placement rules.

Changing Default Storage Class for a User

The default storage class for a placement rule (and therefore a bucket) cannot be modified. If it’s desired that a certain user’s uploads, by default, goto a different storage class, this can be done by modifying the default_storage_class key for the specific user with radosgw-admin .

$ radosgw-admin user info --uid test-user
{
    "user_id": "test-user",
    "display_name": "test-user",
    "email": "",
    "suspended": 0,
    "max_buckets": 1000,
    "subusers": [],
    "keys": [
        {
            "user": "test-user",
            "access_key": "QTAQHWDZBLBSPAK3MEL3",
            "secret_key": "Oe9oqCDY1c3DIgqnK34CBo0M4IWyXkYtTP7MP3Yf"
        }
    ],
    "swift_keys": [],
    "caps": [],
    "op_mask": "read, write, delete",
    "default_placement": "",
    "default_storage_class": "",
    "placement_tags": [],
    "bucket_quota": {
        "enabled": false,
        "check_on_raw": false,
        "max_size": -1,
        "max_size_kb": 0,
        "max_objects": -1
    },
    "user_quota": {
        "enabled": false,
        "check_on_raw": false,
        "max_size": -1,
        "max_size_kb": 0,
        "max_objects": -1
    },
    "temp_url_keys": [],
    "type": "rgw",
    "mfa_ids": []
}
 
$ radosgw-admin user modify \
      --uid test-user \
      --placement-id default-placement \
      --storage-class hot_storage_class
{
    "user_id": "test-user",
    "display_name": "test-user",
    "email": "",
    "suspended": 0,
    "max_buckets": 1000,
    "subusers": [],
    "keys": [
        {
            "user": "test-user",
            "access_key": "QTAQHWDZBLBSPAK3MEL3",
            "secret_key": "Oe9oqCDY1c3DIgqnK34CBo0M4IWyXkYtTP7MP3Yf"
        }
    ],
    "swift_keys": [],
    "caps": [],
    "op_mask": "read, write, delete",
    "default_placement": "default-placement",
    "default_storage_class": "hot_storage_class",
    "placement_tags": [],
    "bucket_quota": {
        "enabled": false,
        "check_on_raw": false,
        "max_size": -1,
        "max_size_kb": 0,
        "max_objects": -1
    },
    "user_quota": {
        "enabled": false,
        "check_on_raw": false,
        "max_size": -1,
        "max_size_kb": 0,
        "max_objects": -1
    },
    "temp_url_keys": [],
    "type": "rgw",
    "mfa_ids": []
}

Otherwise, a storage class can also be specified during the time of the upload like in the placement rule example.

❔ Why?

Storage classes help users and administrators to prioritise space and costs by having different types of storage classes corresponding to different types of device classes. For example, frequently accessed data may reside on NVMe while archives may reside on HDD. This type of transitioning is made possible with lifecycle policies.

🎤 How?

A storage class can be created like so:

Same over the command line:

radosgw-admin zonegroup placement add \
      --rgw-zonegroup default \
      --placement-id default-placement \
      --storage-class hot_storage_class
 
radosgw-admin zone placement add \
      --rgw-zone default \
      --placement-id default-placement \
      --storage-class hot_storage_class \
      --data-pool hot.rgw.data \
      --compression lz4

👓 References

https://docs.ceph.com/en/latest/radosgw/placement/#adding-a-storage-class