Notice

This document is for a development version of Ceph.

Report a Documentation Bug

MDS Config Reference

mds_cache_mid

The insertion point for new items in the cache LRU (from the top).

type:

float

runtime updatable:

true

default:

0.7

mds_allow_batched_ops

The MDS will batch a lookup or getattr RPC on the same inode when possible to avoid repetitive locks on metadata and to bypass other requests acquiring write locks. Generally, this should only improve performance but this switch exists to provide a means to turn this behavior off for comparison.

type:

bool

runtime updatable:

true

default:

true

mds_dir_max_commit_size

The maximum size of a directory update before the MDS splits it into smaller transactions (MB).

type:

int

runtime updatable:

true

default:

10

mds_dir_max_entries

The maximum number of entries before any new entries are rejected with ENOSPC.

type:

uint

runtime updatable:

true

default:

0

mds_decay_halflife

Rate of decay for temperature counters on each directory for balancing

type:

float

runtime updatable:

true

default:

5.0

mds_beacon_interval

Interval in seconds between MDS beacon messages sent to Monitors

type:

float

runtime updatable:

true

default:

4.0

mds_beacon_grace

The interval without beacons before Ceph declares an MDS laggy (and possibly replaces it).

type:

float

runtime updatable:

true

default:

15.0

mon_mds_blocklist_interval

The blocklist duration for failed MDSs in the OSD map. Note, this controls how long failed MDS daemons will stay in the OSDMap blocklist. It has no effect on how long something is blocklisted when the administrator blocklists it manually. For example, ceph osd blocklist add will still use the default blocklist time.

type:

float

runtime updatable:

true

default:

1 day

min:

1_hr

mds_reconnect_timeout

Timeout in seconds to wait for clients to reconnect during MDS reconnect recovery state

type:

float

runtime updatable:

true

default:

45.0

mds_tick_interval

How frequently the MDS performs internal periodic tasks.

type:

float

runtime updatable:

true

default:

5.0

mds_dirstat_min_interval

The minimum interval (in seconds) to try to avoid propagating recursive stats up the tree.

type:

float

runtime updatable:

true

default:

1.0

mds_scatter_nudge_interval

How quickly dirstat changes propagate up.

type:

float

runtime updatable:

true

default:

5.0

mds_client_prealloc_inos

The number of inode numbers to preallocate per client session.

type:

int

runtime updatable:

true

default:

1000

mds_early_reply

Determines whether the MDS should allow clients to see request results before they commit to the journal.

type:

bool

runtime updatable:

true

default:

true

mds_default_dir_hash

The function to use for hashing files across directory fragments.

type:

int

runtime updatable:

true

default:

2

mds_log_skip_corrupt_events

Determines whether the MDS should try to skip corrupt journal events during journal replay.

type:

bool

runtime updatable:

true

default:

false

mds_bal_sample_interval

Determines how frequently to sample directory temperature (for fragmentation decisions).

type:

float

runtime updatable:

true

default:

3.0

mds_bal_replicate_threshold

The minimum temperature before Ceph attempts to replicate metadata to other nodes.

type:

float

runtime updatable:

true

default:

8000.0

mds_bal_unreplicate_threshold

The minimum temperature before Ceph stops replicating metadata to other nodes.

type:

float

runtime updatable:

true

default:

0.0

mds_bal_split_size

The maximum directory size before the MDS will split a directory fragment into smaller bits.

type:

int

runtime updatable:

true

default:

10000

mds_bal_split_rd

The maximum directory read temperature before Ceph splits a directory fragment.

type:

float

runtime updatable:

true

default:

25000.0

mds_bal_split_wr

The maximum directory write temperature before Ceph splits a directory fragment.

type:

float

runtime updatable:

true

default:

10000.0

mds_bal_split_bits

The number of bits by which to split a directory fragment.

type:

int

runtime updatable:

true

default:

3

allowed range:

[1, 24]

mds_bal_merge_size

The minimum directory size before Ceph tries to merge adjacent directory fragments.

type:

int

runtime updatable:

true

default:

50

mds_bal_interval

The frequency (in seconds) of workload exchanges between MDSs.

type:

int

runtime updatable:

true

default:

10

mds_bal_fragment_interval

The delay (in seconds) between a fragment being eligible for split or merge and executing the fragmentation change.

type:

int

runtime updatable:

true

default:

5

mds_bal_fragment_fast_factor

The ratio by which frags may exceed the split size before a split is executed immediately (skipping the fragment interval)

type:

float

runtime updatable:

true

default:

1.5

mds_bal_fragment_size_max

The maximum size of a fragment before any new entries are rejected with ENOSPC.

type:

int

runtime updatable:

true

default:

100000

mds_bal_idle_threshold

The minimum temperature before Ceph migrates a subtree back to its parent.

type:

float

runtime updatable:

true

default:

0.0

mds_bal_max

The number of iterations to run balancer before Ceph stops. (used for testing purposes only)

type:

int

runtime updatable:

true

default:

-1

mds_bal_max_until

The number of seconds to run balancer before Ceph stops. (used for testing purposes only)

type:

int

runtime updatable:

true

default:

-1

mds_bal_mode

The method for calculating MDS load.

• 0 = Hybrid.

• 1 = Request rate and latency.

• 2 = CPU load.

type:

int

runtime updatable:

true

default:

0

mds_bal_min_rebalance

The minimum subtree temperature before Ceph migrates.

type:

float

runtime updatable:

true

default:

0.1

mds_bal_overload_epochs

The number of epochs the overload lasts before Ceph migrates, setting it to a higher value can avoid frequent migrations caused by load fluctuations.

type:

int

runtime updatable:

true

default:

2

mds_bal_min_start

The minimum subtree temperature before Ceph searches a subtree.

type:

float

runtime updatable:

true

default:

0.2

mds_bal_need_min

The minimum fraction of target subtree size to accept.

type:

float

runtime updatable:

true

default:

0.8

mds_bal_need_max

The maximum fraction of target subtree size to accept.

type:

float

runtime updatable:

true

default:

1.2

mds_bal_midchunk

Ceph will migrate any subtree that is larger than this fraction of the target subtree size.

type:

float

runtime updatable:

true

default:

0.3

mds_bal_minchunk

Ceph will ignore any subtree that is smaller than this fraction of the target subtree size.

type:

float

runtime updatable:

true

default:

0.001

mds_replay_interval

The journal poll interval when in standby-replay mode. (“hot standby”)

type:

float

runtime updatable:

true

default:

1.0

mds_shutdown_check

The interval for polling the cache during MDS shutdown.

type:

int

runtime updatable:

true

default:

0

mds_thrash_exports

Ceph will randomly export subtrees between nodes (testing only).

type:

int

runtime updatable:

true

default:

0

mds_thrash_fragments

Ceph will randomly fragment or merge directories.

type:

int

runtime updatable:

true

default:

0

mds_dump_cache_on_map

Ceph will dump the MDS cache contents to a file on each MDSMap.

type:

bool

runtime updatable:

true

default:

false

mds_dump_cache_after_rejoin

Ceph will dump MDS cache contents to a file after rejoining the cache (during recovery).

type:

bool

runtime updatable:

true

default:

false

mds_verify_scatter

Ceph will assert that various scatter/gather invariants are true (developers only).

type:

bool

runtime updatable:

true

default:

false

mds_debug_scatterstat

Ceph will assert that various recursive stat invariants are true (for developers only).

type:

bool

runtime updatable:

true

default:

false

mds_debug_frag

Ceph will verify directory fragmentation invariants when convenient (developers only).

type:

bool

runtime updatable:

true

default:

false

mds_debug_auth_pins

The debug auth pin invariants (for developers only).

type:

bool

runtime updatable:

true

default:

false

mds_debug_subtrees

The debug subtree invariants (for developers only).

type:

bool

runtime updatable:

true

default:

false

mds_kill_mdstable_at

Ceph will inject MDS failure in MDSTable code (for developers only).

type:

int

runtime updatable:

true

default:

0

mds_kill_export_at

Ceph will inject MDS failure in the subtree export code (for developers only).

type:

int

runtime updatable:

true

default:

0

mds_kill_import_at

Ceph will inject MDS failure in the subtree import code (for developers only).

type:

int

runtime updatable:

true

default:

0

mds_kill_link_at

Ceph will inject MDS failure in hard link code (for developers only).

type:

int

runtime updatable:

true

default:

0

mds_kill_rename_at

Ceph will inject MDS failure in the rename code (for developers only).

type:

int

runtime updatable:

true

default:

0

mds_inject_skip_replaying_inotable

MDS will skip replaying the inotable when replaying the journal logs. (for testing only).

type:

bool

runtime updatable:

true

default:

false

mds_kill_after_journal_logs_flushed

The primary MDS will crash just after the mknod/openc journal logs are flushed to the pool. (for testing only).

type:

bool

runtime updatable:

true

default:

false

mds_wipe_sessions

Ceph will delete all client sessions on startup (for testing only).

type:

bool

runtime updatable:

true

default:

false

mds_wipe_ino_prealloc

Ceph will delete ino preallocation metadata on startup (for testing only).

type:

bool

runtime updatable:

true

default:

false

mds_skip_ino

The number of inode numbers to skip on startup (for testing only).

type:

int

runtime updatable:

true

default:

0

mds_min_caps_per_client

minimum number of capabilities a client may hold

type:

uint

runtime updatable:

true

default:

100

mds_symlink_recovery

Stores symlink target on the first data object of symlink file. Allows recover of symlink using recovery tools.

type:

bool

runtime updatable:

true

default:

true

mds_extraordinary_events_dump_interval

Interval in seconds for dumping the recent in-memory logs when there is an extra-ordinary event. The default is 0 (disabled). The log level should be < 10 and the gather level should be >=10 in debug_mds for enabling this option.

type:

secs

runtime updatable:

true

default:

0

allowed range:

[0, 60]

subv_metrics_window_interval

interval in seconds to hold values in sliding window for subvolume metrics

type:

secs

runtime updatable:

true

default:

30

min:

30

Brought to you by the Ceph Foundation

The Ceph Documentation is a community resource funded and hosted by the non-profit Ceph Foundation. If you would like to support this and our other efforts, please consider joining now.