ZFS Tunables for Lustre Object Storage Servers (OSS)
|metaslab_debug_unload||Prevent ZFS from unloading the spacemaps from a metaslab once it is read in||0||1|
|zfs_arc_max||Maximum size of ARC||50% RAM||75% RAM|
|zfs_dirty_data_max||Amount of dirty data on the system; able to absorb more workload variation before throttling||10% RAM||1 - 4GB|
|zfs_vdev_async_write_active_min_dirty_percent||Threshold below which IO scheduler will limit concurrent operations to the minimum. Above this value, concurrent operations increases linearly until the maximum.||30||20|
|zfs_vdev_async_write_min_active||Minimum asynchronous write I/Os active to each device||1||5|
|zfs_vdev_async_write_max_active||Maximum asynchronous write I/Os active to each device||10||10|
|zfs_vdev_sync_read_min_active||Minimum synchronous read I/Os active to each device||16|
|zfs_vdev_sync_read_max_active||Maximum synchronous read I/Os active to each device||16|
|spl_kmem_cache_slab_limit||Objects of spl_kmem_cache_slab_limit or smaller will be allocated using the Linux slab allocator, large objects use the SPL allocator. A cut-off of 16K was determined to be optimal for architectures using 4K pages.||16384||16384|
When dirty data is less than 30% of
zfs_dirty_data_max, ZFS keeps one outstanding write per VDEV. Dirty data will build up very quickly, and because there is only one outstanding write per disk, ZFS will start to delay or even halt writes.
zfs_dirty_data_max parameter should ideally match the backend storage capability. The code simply uses 10% of system memory as the default.
In addition to the kernel module parameters, it is recommended that ZFS compression is also enabled when creating ZFS datasets for OSTs. Creating Lustre Object Storage Services (OSS) provides examples of the commands to create OSTs with compression enabled.