Ceph RADOS Block Device (RBD)

If you use KVM, QEMU or Hyper-V as your hypervisor, you can configure the Compute service to use Ceph RADOS block devices (RBD) for volumes.

Ceph is a massively scalable, open source, distributed storage system. It is comprised of an object store, block store, and a POSIX-compliant distributed file system. The platform can auto-scale to the exabyte level and beyond. It runs on commodity hardware, is self-healing and self-managing, and has no single point of failure. Due to its open-source nature, you can install and use this portable storage platform in public or private clouds.

Ceph architecture

Note

Supported Ceph versions

The current release cycle model for Ceph targets a new release yearly on 1 March, with there being at most two active stable releases at any time.

For a given OpenStack release, Cinder supports the current Ceph active stable releases plus the two prior releases.

For example, at the time of the OpenStack Wallaby release in April 2021, the Ceph active supported releases are Pacific and Octopus. The Cinder Wallaby release therefore supports Ceph Pacific, Octopus, Nautilus, and Mimic.

Additionally, it is expected that the version of the Ceph client available to Cinder or any of its associated libraries (os-brick, cinderlib) is aligned with the Ceph server version. Mixing server and client versions is unsupported and may lead to anomalous behavior.

The minimum requirements for using Ceph with Hyper-V are Ceph Pacific and Windows Server 2016.

RADOS

Ceph is based on Reliable Autonomic Distributed Object Store (RADOS). RADOS distributes objects across the storage cluster and replicates objects for fault tolerance. RADOS contains the following major components:

Object Storage Device (OSD) Daemon

The storage daemon for the RADOS service, which interacts with the OSD (physical or logical storage unit for your data). You must run this daemon on each server in your cluster. For each OSD, you can have an associated hard drive disk. For performance purposes, pool your hard drive disk with raid arrays, or logical volume management (LVM). By default, the following pools are created: data, metadata, and RBD.

Meta-Data Server (MDS)

Stores metadata. MDSs build a POSIX file system on top of objects for Ceph clients. However, if you do not use the Ceph file system, you do not need a metadata server.

Monitor (MON)

A lightweight daemon that handles all communications with external applications and clients. It also provides a consensus for distributed decision making in a Ceph/RADOS cluster. For instance, when you mount a Ceph shared on a client, you point to the address of a MON server. It checks the state and the consistency of the data. In an ideal setup, you must run at least three ceph-mon daemons on separate servers.

Ways to store, use, and expose data

To store and access your data, you can use the following storage systems:

RADOS

Use as an object, default storage mechanism.

RBD

Use as a block device. The Linux kernel RBD (RADOS block device) driver allows striping a Linux block device over multiple distributed object store data objects. It is compatible with the KVM RBD image.

CephFS

Use as a file, POSIX-compliant file system.

Ceph exposes RADOS; you can access it through the following interfaces:

RADOS Gateway

OpenStack Object Storage and Amazon-S3 compatible RESTful interface (see RADOS_Gateway).

librados

and its related C/C++ bindings

RBD and QEMU-RBD

Linux kernel and QEMU block devices that stripe data across multiple objects.

RBD pool

The RBD pool used by the Cinder backend is configured with option rbd_pool, and by default the driver expects exclusive management access to that pool, as in being the only system creating and deleting resources in it, since that’s the recommended deployment choice.

Pool sharing is strongly discouraged, and if we were to share the pool with other services, within OpenStack (Nova, Glance, another Cinder backend) or outside of OpenStack (oVirt), then the stats returned by the driver to the scheduler would not be entirely accurate.

The inaccuracy would be that the actual size in use by the cinder volumes would be lower than the reported one, since it would be also including the used space by the other services.

We can set the rbd_exclusive_cinder_pool configuration option to false to fix this inaccuracy, but this has a performance impact.

Warning

Setting rbd_exclusive_cinder_pool to false will increase the burden on the Cinder driver and the Ceph cluster, since a request will be made for each existing image, to retrieve its size, during the stats gathering process.

For deployments with large amount of volumes it is recommended to leave the default value of true, and accept the inaccuracy, as it should not be particularly problematic.

Driver options

The following table contains the configuration options supported by the Ceph RADOS Block Device driver.

Description of Ceph storage configuration options

Configuration option = Default value	Description
deferred_deletion_delay = 0	(Integer) Time delay in seconds before a volume is eligible for permanent removal after being tagged for deferred deletion.
deferred_deletion_purge_interval = 60	(Integer) Number of seconds between runs of the periodic task to purge volumes tagged for deletion.
enable_deferred_deletion = False	(Boolean) Enable deferred deletion. Upon deletion, volumes are tagged for deletion but will only be removed asynchronously at a later time.
rados_connect_timeout = -1	(Integer) Timeout value (in seconds) used when connecting to ceph cluster. If value < 0, no timeout is set and default librados value is used.
rados_connection_interval = 5	(Integer) Interval value (in seconds) between connection retries to ceph cluster.
rados_connection_retries = 3	(Integer) Number of retries if connection to ceph cluster failed.
rbd_ceph_conf = <>	(String) Path to the ceph configuration file
rbd_cluster_name = ceph	(String) The name of ceph cluster
rbd_concurrent_flatten_operations = 3	(Integer(min=0)) Number of flatten operations that will run concurrently on this volume service.
rbd_exclusive_cinder_pool = True	(Boolean) Set to False if the pool is shared with other usages. On exclusive use driver won’t query images’ provisioned size as they will match the value calculated by the Cinder core code for allocated_capacity_gb. This reduces the load on the Ceph cluster as well as on the volume service. On non exclusive use driver will query the Ceph cluster for per image used disk, this is an intensive operation having an independent request for each image.
rbd_flatten_volume_from_snapshot = False	(Boolean) Flatten volumes created from snapshots to remove dependency from volume to snapshot
rbd_max_clone_depth = 5	(Integer) Maximum number of nested volume clones that are taken before a flatten occurs. Set to 0 to disable cloning. Note: lowering this value will not affect existing volumes whose clone depth exceeds the new value.
rbd_pool = rbd	(String) The RADOS pool where RBD volumes are stored
rbd_secret_uuid = None	(String) The libvirt uuid of the secret for the rbd_user volumes. Defaults to the cluster FSID.
rbd_store_chunk_size = 4	(Integer) Volumes will be chunked into objects of this size (in megabytes).
rbd_user = None	(String) The RADOS client name for accessing RBD volumes - only set when using cephx authentication
replication_connect_timeout = 5	(Integer) Timeout value (in seconds) used when connecting to ceph cluster to do a demotion/promotion of volumes. If value < 0, no timeout is set and default librados value is used.
report_dynamic_total_capacity = True	(Boolean) Set to True for driver to report total capacity as a dynamic value (used + current free) and to False to report a static value (quota max bytes if defined and global size of cluster if not).

RBD Mirroring

The cinder RBD driver supports mirroring between multiple clusters. You can configure it on the cinder side with the usual replication configuration. Refer to the documentation for more information.

You will also have to configure replication on the Ceph side. To do so you may refer to the Ceph documentation.

Note that with the RBD driver in cinder you need to configure the pool replication option in image mode. For instance, if your pool is named volumes, the command would be: rbd mirror pool enable volumes image.

RBD QoS

Currently, the Cinder RBD driver supports the following QoS options compatible with Ceph Octopus release and above:

Cinder Value	Ceph Mapping
total_iops_sec	rbd_qos_iops_limit
read_iops_sec	rbd_qos_read_iops_limit
write_iops_sec	rbd_qos_write_iops_limit
total_bytes_sec	rbd_qos_bps_limit
read_bytes_sec	rbd_qos_read_bps_limit
write_bytes_sec	rbd_qos_write_bps_limit
total_iops_sec_max	rbd_qos_bps_burst
read_iops_sec_max	rbd_qos_read_iops_burst
write_iops_sec_max	rbd_qos_write_iops_burst
total_bytes_sec_max	rbd_qos_bps_burst
read_bytes_sec_max	rbd_qos_read_bps_burst
write_bytes_sec_max	rbd_qos_write_bps_burst

For more information on QoS settings you may refer to Ceph QoS documentation.