Struct PartitionCreateOptions

pub struct PartitionCreateOptions { /* private fields */ }

Options to configure a partition

Implementations

impl CreateOptions

pub fn compression(self, compression: CompressionType) -> Self

Sets the compression method.

Once set for a partition, this property is not considered in the future.

Default = In order: Lz4 -> Miniz -> None, depending on compilation flags

pub fn compaction_strategy( self, compaction_strategy: CompactionStrategy, ) -> Self

Sets the compaction strategy.

Default = Leveled

pub fn manual_journal_persist(self, flag: bool) -> Self

If false, writes will flush data to the operating system.

Default = false

Set to true to handle persistence manually, e.g. manually using PersistMode::SyncData.

pub fn max_memtable_size(self, bytes: u32) -> Self

Sets the maximum memtable size.

Default = 16 MiB

Recommended size 8 - 64 MiB, depending on how much memory is available.

Note that the memory usage may temporarily be max_memtable_size * flush_worker_count because of parallel flushing. Use the keyspace’s max_write_buffer_size to cap global memory usage.

Conversely, if max_memtable_size is larger than 64 MiB, it may require increasing the keyspace’s max_write_buffer_size.

pub fn block_size(self, block_size: u32) -> Self

Sets the block size.

Once set for a partition, this property is not considered in the future.

Default = 4 KiB

For point read heavy workloads (get) a sensible default is somewhere between 4 - 8 KiB, depending on the average value size.

For scan heavy workloads (range, prefix), use 16 - 64 KiB which also increases compression efficiency.

Panics

Panics if the block size is smaller than 1 KiB or larger than 512 KiB.

pub fn with_kv_separation(self, opts: KvSeparationOptions) -> Self

Enables key-value separation for this partition.

Key-value separation is intended for large value scenarios (1 KiB+ per KV). Large values will be separated into a log-structured value log, which heavily decreases compaction overhead at the cost of slightly higher read latency and higher temporary space usage. Also, garbage collection for deleted or outdated values becomes lazy, so GC needs to be triggered manually. See https://fjall-rs.github.io/post/announcing-fjall-2/#key-value-separation for more information.

Once set for a partition, this property is not considered in the future.

Default = disabled

Trait Implementations

impl Clone for CreateOptions

fn clone(&self) -> CreateOptions

Returns a copy of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for CreateOptions

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Decode for CreateOptions

fn decode_from<R: Read>(reader: &mut R) -> Result<Self, DecodeError>

where Self: Sized,

Deserializes from reader.

impl Default for CreateOptions

fn default() -> Self

Returns the “default value” for a type.

impl Encode for CreateOptions

fn encode_into<W: Write>(&self, writer: &mut W) -> Result<(), EncodeError>

Serializes into writer.

fn encode_into_vec(&self) -> Vec<u8>

Serializes into vector.