========================== BlueStore Config Reference ========================== Devices ======= BlueStore manages either one, two, or (in certain cases) three storage devices. In the simplest case, BlueStore consumes a single (primary) storage device. The storage device is normally used as a whole, occupying the full device that is managed directly by BlueStore. This *primary device* is normally identified by a ``block`` symlink in the data directory. The data directory is a ``tmpfs`` mount which gets populated (at boot time, or when ``ceph-volume`` activates it) with all the common OSD files that hold information about the OSD, like: its identifier, which cluster it belongs to, and its private keyring. It is also possible to deploy BlueStore across one or two additional devices: * A *write-ahead log (WAL) device* (identified as ``block.wal`` in the data directory) can be used for BlueStore's internal journal or write-ahead log. It is only useful to use a WAL device if the device is faster than the primary device (e.g., when it is on an SSD and the primary device is an HDD). * A *DB device* (identified as ``block.db`` in the data directory) can be used for storing BlueStore's internal metadata. BlueStore (or rather, the embedded RocksDB) will put as much metadata as it can on the DB device to improve performance. If the DB device fills up, metadata will spill back onto the primary device (where it would have been otherwise). Again, it is only helpful to provision a DB device if it is faster than the primary device. If there is only a small amount of fast storage available (e.g., less than a gigabyte), we recommend using it as a WAL device. If there is more, provisioning a DB device makes more sense. The BlueStore journal will always be placed on the fastest device available, so using a DB device will provide the same benefit that the WAL device would while *also* allowing additional metadata to be stored there (if it will fit). This means that if a DB device is specified but an explicit WAL device is not, the WAL will be implicitly colocated with the DB on the faster device. A single-device (colocated) BlueStore OSD can be provisioned with: .. prompt:: bash $ ceph-volume lvm prepare --bluestore --data To specify a WAL device and/or DB device: .. prompt:: bash $ ceph-volume lvm prepare --bluestore --data --block.wal --block.db .. note:: ``--data`` can be a Logical Volume using *vg/lv* notation. Other devices can be existing logical volumes or GPT partitions. Provisioning strategies ----------------------- Although there are multiple ways to deploy a BlueStore OSD (unlike Filestore which had just one), there are two common arrangements that should help clarify the deployment strategy: .. _bluestore-single-type-device-config: **block (data) only** ^^^^^^^^^^^^^^^^^^^^^ If all devices are the same type, for example all rotational drives, and there are no fast devices to use for metadata, it makes sense to specify the block device only and to not separate ``block.db`` or ``block.wal``. The :ref:`ceph-volume-lvm` command for a single ``/dev/sda`` device looks like: .. prompt:: bash $ ceph-volume lvm create --bluestore --data /dev/sda If logical volumes have already been created for each device, (a single LV using 100% of the device), then the :ref:`ceph-volume-lvm` call for an LV named ``ceph-vg/block-lv`` would look like: .. prompt:: bash $ ceph-volume lvm create --bluestore --data ceph-vg/block-lv .. _bluestore-mixed-device-config: **block and block.db** ^^^^^^^^^^^^^^^^^^^^^^ If you have a mix of fast and slow devices (SSD / NVMe and rotational), it is recommended to place ``block.db`` on the faster device while ``block`` (data) lives on the slower (spinning drive). You must create these volume groups and logical volumes manually as the ``ceph-volume`` tool is currently not able to do so automatically. For the below example, let us assume four rotational (``sda``, ``sdb``, ``sdc``, and ``sdd``) and one (fast) solid state drive (``sdx``). First create the volume groups: .. prompt:: bash $ vgcreate ceph-block-0 /dev/sda vgcreate ceph-block-1 /dev/sdb vgcreate ceph-block-2 /dev/sdc vgcreate ceph-block-3 /dev/sdd Now create the logical volumes for ``block``: .. prompt:: bash $ lvcreate -l 100%FREE -n block-0 ceph-block-0 lvcreate -l 100%FREE -n block-1 ceph-block-1 lvcreate -l 100%FREE -n block-2 ceph-block-2 lvcreate -l 100%FREE -n block-3 ceph-block-3 We are creating 4 OSDs for the four slow spinning devices, so assuming a 200GB SSD in ``/dev/sdx`` we will create 4 logical volumes, each of 50GB: .. prompt:: bash $ vgcreate ceph-db-0 /dev/sdx lvcreate -L 50GB -n db-0 ceph-db-0 lvcreate -L 50GB -n db-1 ceph-db-0 lvcreate -L 50GB -n db-2 ceph-db-0 lvcreate -L 50GB -n db-3 ceph-db-0 Finally, create the 4 OSDs with ``ceph-volume``: .. prompt:: bash $ ceph-volume lvm create --bluestore --data ceph-block-0/block-0 --block.db ceph-db-0/db-0 ceph-volume lvm create --bluestore --data ceph-block-1/block-1 --block.db ceph-db-0/db-1 ceph-volume lvm create --bluestore --data ceph-block-2/block-2 --block.db ceph-db-0/db-2 ceph-volume lvm create --bluestore --data ceph-block-3/block-3 --block.db ceph-db-0/db-3 These operations should end up creating four OSDs, with ``block`` on the slower rotational drives with a 50 GB logical volume (DB) for each on the solid state drive. Sizing ====== When using a :ref:`mixed spinning and solid drive setup ` it is important to make a large enough ``block.db`` logical volume for BlueStore. Generally, ``block.db`` should have *as large as possible* logical volumes. The general recommendation is to have ``block.db`` size in between 1% to 4% of ``block`` size. For RGW workloads, it is recommended that the ``block.db`` size isn't smaller than 4% of ``block``, because RGW heavily uses it to store metadata (omap keys). For example, if the ``block`` size is 1TB, then ``block.db`` shouldn't be less than 40GB. For RBD workloads, 1% to 2% of ``block`` size is usually enough. In older releases, internal level sizes mean that the DB can fully utilize only specific partition / LV sizes that correspond to sums of L0, L0+L1, L1+L2, etc. sizes, which with default settings means roughly 3 GB, 30 GB, 300 GB, and so forth. Most deployments will not substantially benefit from sizing to accommodate L3 and higher, though DB compaction can be facilitated by doubling these figures to 6GB, 60GB, and 600GB. Improvements in releases beginning with Nautilus 14.2.12 and Octopus 15.2.6 enable better utilization of arbitrary DB device sizes, and the Pacific release brings experimental dynamic level support. Users of older releases may thus wish to plan ahead by provisioning larger DB devices today so that their benefits may be realized with future upgrades. When *not* using a mix of fast and slow devices, it isn't required to create separate logical volumes for ``block.db`` (or ``block.wal``). BlueStore will automatically colocate these within the space of ``block``. Automatic Cache Sizing ====================== BlueStore can be configured to automatically resize its caches when TCMalloc is configured as the memory allocator and the ``bluestore_cache_autotune`` setting is enabled. This option is currently enabled by default. BlueStore will attempt to keep OSD heap memory usage under a designated target size via the ``osd_memory_target`` configuration option. This is a best effort algorithm and caches will not shrink smaller than the amount specified by ``osd_memory_cache_min``. Cache ratios will be chosen based on a hierarchy of priorities. If priority information is not available, the ``bluestore_cache_meta_ratio`` and ``bluestore_cache_kv_ratio`` options are used as fallbacks. Manual Cache Sizing =================== The amount of memory consumed by each OSD for BlueStore caches is determined by the ``bluestore_cache_size`` configuration option. If that config option is not set (i.e., remains at 0), there is a different default value that is used depending on whether an HDD or SSD is used for the primary device (set by the ``bluestore_cache_size_ssd`` and ``bluestore_cache_size_hdd`` config options). BlueStore and the rest of the Ceph OSD daemon do the best they can to work within this memory budget. Note that on top of the configured cache size, there is also memory consumed by the OSD itself, and some additional utilization due to memory fragmentation and other allocator overhead. The configured cache memory budget can be used in a few different ways: * Key/Value metadata (i.e., RocksDB's internal cache) * BlueStore metadata * BlueStore data (i.e., recently read or written object data) Cache memory usage is governed by the following options: ``bluestore_cache_meta_ratio`` and ``bluestore_cache_kv_ratio``. The fraction of the cache devoted to data is governed by the effective bluestore cache size (depending on ``bluestore_cache_size[_ssd|_hdd]`` settings and the device class of the primary device) as well as the meta and kv ratios. The data fraction can be calculated by `` * (1 - bluestore_cache_meta_ratio - bluestore_cache_kv_ratio)`` Checksums ========= BlueStore checksums all metadata and data written to disk. Metadata checksumming is handled by RocksDB and uses `crc32c`. Data checksumming is done by BlueStore and can make use of `crc32c`, `xxhash32`, or `xxhash64`. The default is `crc32c` and should be suitable for most purposes. Full data checksumming does increase the amount of metadata that BlueStore must store and manage. When possible, e.g., when clients hint that data is written and read sequentially, BlueStore will checksum larger blocks, but in many cases it must store a checksum value (usually 4 bytes) for every 4 kilobyte block of data. It is possible to use a smaller checksum value by truncating the checksum to two or one byte, reducing the metadata overhead. The trade-off is that the probability that a random error will not be detected is higher with a smaller checksum, going from about one in four billion with a 32-bit (4 byte) checksum to one in 65,536 for a 16-bit (2 byte) checksum or one in 256 for an 8-bit (1 byte) checksum. The smaller checksum values can be used by selecting `crc32c_16` or `crc32c_8` as the checksum algorithm. The *checksum algorithm* can be set either via a per-pool ``csum_type`` property or the global config option. For example: .. prompt:: bash $ ceph osd pool set csum_type Inline Compression ================== BlueStore supports inline compression using `snappy`, `zlib`, or `lz4`. Please note that the `lz4` compression plugin is not distributed in the official release. Whether data in BlueStore is compressed is determined by a combination of the *compression mode* and any hints associated with a write operation. The modes are: * **none**: Never compress data. * **passive**: Do not compress data unless the write operation has a *compressible* hint set. * **aggressive**: Compress data unless the write operation has an *incompressible* hint set. * **force**: Try to compress data no matter what. For more information about the *compressible* and *incompressible* IO hints, see :c:func:`rados_set_alloc_hint`. Note that regardless of the mode, if the size of the data chunk is not reduced sufficiently it will not be used and the original (uncompressed) data will be stored. For example, if the ``bluestore compression required ratio`` is set to ``.7`` then the compressed data must be 70% of the size of the original (or smaller). The *compression mode*, *compression algorithm*, *compression required ratio*, *min blob size*, and *max blob size* can be set either via a per-pool property or a global config option. Pool properties can be set with: .. prompt:: bash $ ceph osd pool set compression_algorithm ceph osd pool set compression_mode ceph osd pool set compression_required_ratio ceph osd pool set compression_min_blob_size ceph osd pool set compression_max_blob_size .. _bluestore-rocksdb-sharding: RocksDB Sharding ================ Internally BlueStore uses multiple types of key-value data, stored in RocksDB. Each data type in BlueStore is assigned a unique prefix. Until Pacific all key-value data was stored in single RocksDB column family: 'default'. Since Pacific, BlueStore can divide this data into multiple RocksDB column families. When keys have similar access frequency, modification frequency and lifetime, BlueStore benefits from better caching and more precise compaction. This improves performance, and also requires less disk space during compaction, since each column family is smaller and can compact independent of others. OSDs deployed in Pacific or later use RocksDB sharding by default. If Ceph is upgraded to Pacific from a previous version, sharding is off. To enable sharding and apply the Pacific defaults, stop an OSD and run .. prompt:: bash # ceph-bluestore-tool \ --path \ --sharding="m(3) p(3,0-12) O(3,0-13)=block_cache={type=binned_lru} L P" \ reshard Throttling ========== SPDK Usage ================== If you want to use the SPDK driver for NVMe devices, you must prepare your system. Refer to `SPDK document`__ for more details. .. __: http://www.spdk.io/doc/getting_started.html#getting_started_examples SPDK offers a script to configure the device automatically. Users can run the script as root: .. prompt:: bash $ sudo src/spdk/scripts/setup.sh You will need to specify the subject NVMe device's device selector with the "spdk:" prefix for ``bluestore_block_path``. For example, you can find the device selector of an Intel PCIe SSD with: .. prompt:: bash $ lspci -mm -n -D -d 8086:0953 The device selector always has the form of ``DDDD:BB:DD.FF`` or ``DDDD.BB.DD.FF``. and then set:: bluestore_block_path = "spdk:trtype:PCIe traddr:0000:01:00.0" Where ``0000:01:00.0`` is the device selector found in the output of ``lspci`` command above. You may also specify a remote NVMeoF target over the TCP transport as in the following example:: bluestore_block_path = "spdk:trtype:TCP traddr:10.67.110.197 trsvcid:4420 subnqn:nqn.2019-02.io.spdk:cnode1" To run multiple SPDK instances per node, you must specify the amount of dpdk memory in MB that each instance will use, to make sure each instance uses its own DPDK memory. In most cases, a single device can be used for data, DB, and WAL. We describe this strategy as *colocating* these components. Be sure to enter the below settings to ensure that all IOs are issued through SPDK.:: bluestore_block_db_path = "" bluestore_block_db_size = 0 bluestore_block_wal_path = "" bluestore_block_wal_size = 0 Otherwise, the current implementation will populate the SPDK map files with kernel file system symbols and will use the kernel driver to issue DB/WAL IO. Minimum Allocation Size ======================== There is a configured minimum amount of storage that BlueStore will allocate on an OSD. In practice, this is the least amount of capacity that a RADOS object can consume. The value of `bluestore_min_alloc_size` is derived from the value of `bluestore_min_alloc_size_hdd` or `bluestore_min_alloc_size_ssd` depending on the OSD's ``rotational`` attribute. This means that when an OSD is created on an HDD, BlueStore will be initialized with the current value of `bluestore_min_alloc_size_hdd`, and SSD OSDs (including NVMe devices) with the value of `bluestore_min_alloc_size_ssd`. Through the Mimic release, the default values were 64KB and 16KB for rotational (HDD) and non-rotational (SSD) media respectively. Octopus changed the default for SSD (non-rotational) media to 4KB, and Pacific changed the default for HDD (rotational) media to 4KB as well. These changes were driven by space amplification experienced by Ceph RADOS GateWay (RGW) deployments that host large numbers of small files (S3/Swift objects). For example, when an RGW client stores a 1KB S3 object, it is written to a single RADOS object. With the default `min_alloc_size` value, 4KB of underlying drive space is allocated. This means that roughly (4KB - 1KB) == 3KB is allocated but never used, which corresponds to 300% overhead or 25% efficiency. Similarly, a 5KB user object will be stored as one 4KB and one 1KB RADOS object, again stranding 4KB of device capcity, though in this case the overhead is a much smaller percentage. Think of this in terms of the remainder from a modulus operation. The overhead *percentage* thus decreases rapidly as user object size increases. An easily missed additional subtlety is that this takes place for *each* replica. So when using the default three copies of data (3R), a 1KB S3 object actually consumes roughly 9KB of storage device capacity. If erasure coding (EC) is used instead of replication, the amplification may be even higher: for a ``k=4,m=2`` pool, our 1KB S3 object will allocate (6 * 4KB) = 24KB of device capacity. When an RGW bucket pool contains many relatively large user objects, the effect of this phenomenon is often negligible, but should be considered for deployments that expect a signficiant fraction of relatively small objects. The 4KB default value aligns well with conventional HDD and SSD devices. Some new coarse-IU (Indirection Unit) QLC SSDs however perform and wear best when `bluestore_min_alloc_size_ssd` is set at OSD creation to match the device's IU:. 8KB, 16KB, or even 64KB. These novel storage drives allow one to achieve read performance competitive with conventional TLC SSDs and write performance faster than HDDs, with high density and lower cost than TLC SSDs. Note that when creating OSDs on these devices, one must carefully apply the non-default value only to appropriate devices, and not to conventional SSD and HDD devices. This may be done through careful ordering of OSD creation, custom OSD device classes, and especially by the use of central configuration _masks_. Quincy and later releases add the `bluestore_use_optimal_io_size_for_min_alloc_size` option that enables automatic discovery of the appropriate value as each OSD is created. Note that the use of ``bcache``, ``OpenCAS``, ``dmcrypt``, ``ATA over Ethernet``, `iSCSI`, or other device layering / abstraction technologies may confound the determination of appropriate values. OSDs deployed on top of VMware storage have been reported to also sometimes report a ``rotational`` attribute that does not match the underlying hardware. We suggest inspecting such OSDs at startup via logs and admin sockets to ensure that behavior is appropriate. Note that this also may not work as desired with older kernels. You can check for this by examining the presence and value of ``/sys/block//queue/optimal_io_size``. You may also inspect a given OSD: .. prompt:: bash # ceph osd metadata osd.1701 | grep rotational This space amplification may manifest as an unusually high ratio of raw to stored data reported by ``ceph df``. ``ceph osd df`` may also report anomalously high ``%USE`` / ``VAR`` values when compared to other, ostensibly identical OSDs. A pool using OSDs with mismatched ``min_alloc_size`` values may experience unexpected balancer behavior as well. Note that this BlueStore attribute takes effect *only* at OSD creation; if changed later, a given OSD's behavior will not change unless / until it is destroyed and redeployed with the appropriate option value(s). Upgrading to a later Ceph release will *not* change the value used by OSDs deployed under older releases or with other settings. DSA (Data Streaming Accelerator Usage) ====================================== If you want to use the DML library to drive DSA device for offloading read/write operations on Persist memory in Bluestore. You need to install `DML`_ and `idxd-config`_ library in your machine with SPR (Sapphire Rapids) CPU. .. _DML: https://github.com/intel/DML .. _idxd-config: https://github.com/intel/idxd-config After installing the DML software, you need to configure the shared work queues (WQs) with the following WQ configuration example via accel-config tool: .. prompt:: bash $ accel-config config-wq --group-id=1 --mode=shared --wq-size=16 --threshold=15 --type=user --name="MyApp1" --priority=10 --block-on-fault=1 dsa0/wq0.1 accel-config config-engine dsa0/engine0.1 --group-id=1 accel-config enable-device dsa0 accel-config enable-wq dsa0/wq0.1