Configuring Metadata Server daemons
Use this list of commands for Metadata Server (MDS) daemon configuration.
| Command | Description | Type | Default |
|---|---|---|---|
| mon_force_standby_active | If set to true, monitors force MDS in standby replay mode to be active. Set under the [mon] or [global] section in the Ceph configuration file. |
Boolean | true |
| max_mds | The number of active MDS daemons during cluster creation. Set under the [mon] or [global] section in the Ceph configuration file. |
32-bit Integer | 1 |
| mds_cache_memory_limit | The memory limit that the MDS enforces for its cache. Use this parameter instead of the MDS cache size parameter. | 64-bit Integer Unsigned | 1073741824 |
| mds_cache_reservation | The cache reservation, memory or inodes, for the MDS cache to maintain. The value is a percentage of the maximum cache configured. Once the MDS begins dipping into its reservation, it recalls client state until its cache size shrinks to restore the reservation. | Float | 0.05 |
| mds_cache_size | The number of inodes to cache. A value of 0 indicates an unlimited number. Red Hat recommends to use the mds_cache_memory_limit to limit the amount of memory the MDS cache uses. |
32-bit Integer | 0 |
| mds_cache_mid | The insertion point for new items in the cache LRU, from the top. | Float | 0.7 |
| mds_dir_commit_ratio | The fraction of directory that contains erroneous information before Ceph commits using a full update instead of partial update. | Float | 0.5 |
| mds_dir_max_commit_size | The maximum size of a directory update in MB before Ceph breaks the directory into smaller transactions. | 32-bit Integer | 10 |
| mds_decay_halflife | The half-life of the MDS cache temperature. | Float | 5 |
| mds_beacon_interval | The frequency, in seconds, of beacon messages sent to the monitor. | Float | 4 |
| mds_beacon_grace | The interval without beacons before Ceph declares a MDS laggy and possibly replaces it. |
Float | 15 |
| mds_blacklist_interval | The blacklist duration for failed MDS daemons in the OSD map. | Float | 24.0*60.0 |
| mds_session_timeout | The interval, in seconds, of client inactivity before Ceph times out capabilities and leases. | Float | 60 |
| mds_session_autoclose | The interval, in seconds, before Ceph closes a laggy client’s session. |
Float | 300 |
| mds_reconnect_timeout | The interval, in seconds, to wait for clients to reconnect during a MDS restart. | Float | 45 |
| mds_tick_interval | How frequently the MDS performs internal periodic tasks. | Float | 5 |
| mds_dirstat_min_interval | The minimum interval, in seconds, to try to avoid propagating recursive statistics up the tree. | Float | 1 |
| mds_scatter_nudge_interval | How quickly changes in directory statistics propagate up. | Float | 5 |
| mds_client_prealloc_inos | The number of inode numbers to preallocate per client session. | 32-bit Integer | 1000 |
| mds_early_reply | Determines whether the MDS allows clients to see request results before they commit to the journal. | Boolean | true |
| mds_use_tmap | Use trivialmap for directory updates. |
Boolean | true |
| mds_default_dir_hash | The function to use for hashing files across directory fragments. | 32-bit Integer | 2, that is, rjenkins |
| mds_log | Set to true if the MDS should journal metadata updates. Disable for benchmarking only. |
Boolean | true |
| mds_log_skip_corrupt_events | Determines whether the MDS tries to skip corrupt journal events during journal replay. | Boolean | false |
| mds_log_max_events | The maximum events in the journal before Ceph initiates trimming. Set to -1 to disable limits. |
32-bit Integer | -1 |
| mds_log_max_segments | The maximum number of segments or objects in the journal before Ceph initiates trimming. Set to -1 to disable limits. |
32-bit Integer | 30 |
| mds_log_max_expiring | The maximum number of segments to expire in parallels. | 32-bit Integer | 20 |
| mds_log_eopen_size | The maximum number of inodes in an EOpen event. |
32-bit Integer | 100 |
| mds_bal_sample_interval | Determines how frequently to sample directory temperature when making fragmentation decisions. | Float | 3 |
| mds_bal_replicate_threshold | The maximum temperature before Ceph attempts to replicate metadata to other nodes. | Float | 8000 |
| mds_bal_unreplicate_threshold | The minimum temperature before Ceph stops replicating metadata to other nodes. | Float | 0 |
| mds_bal_frag | Determines whether or not the MDS fragments directories. | Boolean | false |
| mds_bal_split_size | The maximum directory size before the MDS splits a directory fragment into smaller bits. The root directory has a default fragment size limit of 10000. | 32-bit Integer | 10000 |
| mds_bal_split_rd | The maximum directory read temperature before Ceph splits a directory fragment. | Float | 25000 |
| mds_bal_split_wr | The maximum directory write temperature before Ceph splits a directory fragment. | Float | 10000 |
| mds_bal_split_bits | The number of bits by which to split a directory fragment. | 32-bit Integer | 3 |
| mds_bal_merge_size | The minimum directory size before Ceph tries to merge adjacent directory fragments. | 32-bit Integer | 50 |
| mds_bal_merge_rd | The minimum read temperature before Ceph merges adjacent directory fragments. | Float | 1000 |
| mds_bal_merge_wr | The minimum write temperature before Ceph merges adjacent directory fragments. | Float | 1000 |
| mds_bal_interval | The frequency, in seconds, of workload exchanges between MDS nodes. | 32-bit Integer | 10 |
| mds_bal_fragment_interval | The frequency, in seconds, of adjusting directory fragmentation. | 32-bit Integer | 5 |
| mds_bal_idle_threshold | The minimum temperature before Ceph migrates a subtree back to its parent. | Float | 0 |
| mds_bal_max | The number of iterations to run balancer before Ceph stops. For testing purposes only. | 32-bit Integer | -1 |
| mds_bal_max_until | The number of seconds to run balancer before Ceph stops. For testing purposes only. | 32-bit Integer | -1 |
| mds_bal_mode | The method for calculating MDS load: | 32-bit Integer | 0 |
| mds_bal_min_rebalance | The minimum subtree temperature before Ceph migrates. | Float | 0.1 |
| mds_bal_min_start | The minimum subtree temperature before Ceph searches a subtree. | Float | 0.2 |
| mds_bal_need_min | The minimum fraction of target subtree size to accept. | Float | 0.8 |
| mds_bal_need_max | The maximum fraction of target subtree size to accept. | Float | 1.2 |
| mds_bal_midchunk | Ceph migrates any subtree that is larger than this fraction of the target subtree size. | Float | 0.3 |
| mds_bal_minchunk | Ceph ignores any subtree that is smaller than this fraction of the target subtree size. | Float | 0.001 |
| mds_bal_target_removal_min | The minimum number of balancer iterations before Ceph removes an old MDS target from the MDS map. | 32-bit Integer | 5 |
| mds_bal_target_removal_max | The maximum number of balancer iterations before Ceph removes an old MDS target from the MDS map. | 32-bit Integer | 10 |
| mds_replay_interval | The journal poll interval when in standby-replay mode for a hot standby. |
Float | 1 |
| mds_shutdown_check | The interval for polling the cache during MDS shutdown. | 32-bit Integer | 0 |
| mds_thrash_exports | Ceph randomly exports subtrees between nodes. For testing purposes only. | 32-bit Integer | 0 |
| mds_thrash_fragments | Ceph randomly fragments or merges directories. | 32-bit Integer | 0 |
| mds_dump_cache_on_map | Ceph dumps the MDS cache contents to a file on each MDS map. | Boolean | false |
| mds_dump_cache_after_rejoin | Ceph dumps MDS cache contents to a file after rejoining the cache during recovery. | Boolean | false |
| mds_verify_scatter | Ceph asserts that various scatter/gather invariants are true. For developer use only. |
Boolean | false |
| mds_debug_scatterstat | Ceph asserts that various recursive statistics invariants are true. For developer use only. |
Boolean | false |
| mds_debug_frag | Ceph verifies directory fragmentation invariants when convenient. For developer use only. | Boolean | false |
| mds_debug_auth_pins | The debug authentication pin invariants. For developer use only. | Boolean | false |
| mds_debug_subtrees | Debugging subtree invariants. For developer use only. | Boolean | false |
| mds_kill_mdstable_at | Ceph injects a MDS failure in a MDS Table code. For developer use only. | 32-bit Integer | 0 |
| mds_kill_export_at | Ceph injects a MDS failure in the subtree export code. For developer use only. | 32-bit Integer | 0 |
| mds_kill_import_at | Ceph injects a MDS failure in the subtree import code. For developer use only. | 32-bit Integer | 0 |
| mds_kill_link_at | Ceph injects a MDS failure in a hard link code. For developer use only. | 32-bit Integer | 0 |
| mds_kill_rename_at | Ceph injects a MDS failure in the rename code. For developer use only. | 32-bit Integer | 0 |
| mds_wipe_sessions | Ceph deletes all client sessions on startup. For testing purposes only. | Boolean | 0 |
| mds_wipe_ino_prealloc | Ceph deletes inode preallocation metadata on startup. For testing purposes only. | Boolean | 0 |
| mds_skip_ino | The number of inode numbers to skip on startup. For testing purposes only. | 32-bit Integer | 0 |
| mds_standby_for_name | The MDS daemon is a standby for another MDS daemon of the name specified in this setting. | String | N/A |
| mds_standby_for_rank | An instance of the MDS daemon is a standby for another MDS daemon instance of this rank. | 32-bit Integer | -1 |
| mds_standby_replay | Determines whether the MDS daemon polls and replays the log of an active MDS when used as a hot standby. |
Boolean | false |