This document is for a development version of Ceph.
ceph-bluestore-tool – bluestore administrative tool¶
ceph-bluestore-tool is a utility to perform low-level administrative operations on a BlueStore instance.
fsck [ –deep ]
run consistency check on BlueStore metadata. If –deep is specified, also read all object data and verify checksums.
Run a consistency check and repair any errors we can.
Export the contents of BlueFS (i.e., RocksDB files) to an output directory.
bluefs-bdev-sizes –path osd path
Print the device sizes, as understood by BlueFS, to stdout.
bluefs-bdev-expand –path osd path
Instruct BlueFS to check the size of its block devices and, if they have expanded, make use of the additional space. Please note that only the new files created by BlueFS will be allocated on the preferred block device if it has enough free space, and the existing files that have spilled over to the slow device will be gradually removed when RocksDB performs compaction. In other words, if there is any data spilled over to the slow device, it will be moved to the fast device over time.
bluefs-bdev-new-wal –path osd path –dev-target new-device
Adds WAL device to BlueFS, fails if WAL device already exists.
bluefs-bdev-new-db –path osd path –dev-target new-device
Adds DB device to BlueFS, fails if DB device already exists.
bluefs-bdev-migrate –dev-target new-device –devs-source device1 [–devs-source device2]
Moves BlueFS data from source device(s) to the target one, source devices (except the main one) are removed on success. Target device can be both already attached or new device. In the latter case it’s added to OSD replacing one of the source devices. Following replacement rules apply (in the order of precedence, stop on the first match):
if source list has DB volume - target device replaces it.
if source list has WAL volume - target device replace it.
if source list has slow volume only - operation isn’t permitted, requires explicit allocation via new-db/new-wal command.
show-label –dev device […]
Show device label(s).
free-dump –path osd path [ –allocator block/bluefs-wal/bluefs-db/bluefs-slow ]
Dump all free regions in allocator.
free-score –path osd path [ –allocator block/bluefs-wal/bluefs-db/bluefs-slow ]
Give a [0-1] number that represents quality of fragmentation in allocator. 0 represents case when all free space is in one chunk. 1 represents worst possible fragmentation.
reshard –path osd path –sharding new sharding [ –resharding-ctrl control string ]
Changes sharding of BlueStore’s RocksDB. Sharding is build on top of RocksDB column families. This option allows to test performance of new sharding without need to redeploy OSD. Resharding is usually a long process, which involves walking through entire RocksDB key space and moving some of them to different column families. Option –resharding-ctrl provides performance control over resharding process. Interrupted resharding will prevent OSD from running. Interrupted resharding does not corrupt data. It is always possible to continue previous resharding, or select any other sharding scheme, including reverting to original one.
show-sharding –path osd path
Show sharding that is currently applied to BlueStore’s RocksDB.
Add device to the list of devices to consider
Add device to the list of devices to consider as sources for migrate operation
Specify target device migrate operation or device to add for adding new DB/WAL.
Specify an osd path. In most cases, the device list is inferred from the symlinks present in osd path. This is usually simpler than explicitly specifying the device(s) with –dev.
Output directory for bluefs-export
file to log to
debug log level. Default is 30 (extremely verbose), 20 is very verbose, 10 is verbose, and 1 is not very verbose.
deep scrub/repair (read and validate object data, not just metadata)
Useful for free-dump and free-score actions. Selects allocator(s).
Provides control over resharding process. Specifies how often refresh RocksDB iterator, and how large should commit batch be before committing to RocksDB. Option format is: <iterator_refresh_bytes>/<iterator_refresh_keys>/<batch_commit_bytes>/<batch_commit_keys> Default: 10000000/10000/1000000/1000
Every BlueStore block device has a single block label at the beginning of the device. You can dump the contents of the label with:
ceph-bluestore-tool show-label --dev *device*
The main device will have a lot of metadata, including information that used to be stored in small files in the OSD data directory. The auxiliary devices (db and wal) will only have the minimum required fields (OSD UUID, size, device type, birth time).
OSD directory priming¶
You can generate the content for an OSD data directory that can start up a BlueStore OSD with the prime-osd-dir command:
ceph-bluestore-tool prime-osd-dir --dev *main device* --path /var/lib/ceph/osd/ceph-*id*
BlueFS log rescue¶
Some versions of BlueStore were susceptible to BlueFS log growing extremaly large - beyond the point of making booting OSD impossible. This state is indicated by booting that takes very long and fails in _replay function.
- This can be fixed by::
ceph-bluestore-tool fsck –path osd path –bluefs_replay_recovery=true
- It is advised to first check if rescue process would be successfull::
ceph-bluestore-tool fsck –path osd path –bluefs_replay_recovery=true –bluefs_replay_recovery_disable_compact=true
If above fsck is successful fix procedure can be applied.
ceph-bluestore-tool is part of Ceph, a massively scalable, open-source, distributed storage system. Please refer to the Ceph documentation at http://ceph.com/docs for more information.