cassandra.yaml configuration file

The cassandra.yaml file is the main configuration file for DataStax Distribution of Apache Cassandra 3.11.

cassandra-rackdc.properties

The cassandra-rackdc.properties file is located in the installation_location/conf directory.

cassandra.yaml

The cassandra.yaml file is located in the installation_location/conf directory.

cassandra-topology.properties

The cassandra-topology.properties file is located in the installation_location/conf directory.

The cassandra.yaml file is the main configuration file for DataStax Distribution of Apache Cassandra™ (DDAC).

Important: After changing properties in the cassandra.yaml file, you must restart the node for the changes to take effect.

The cassandra.yaml file is located in the installation_location/conf directory.

Syntax

For the properties in each section, the main setting has zero spaces, and at least two spaces are required before each entry in that section. Adhere to the YAML syntax.

Organization

The configuration properties are grouped into the following sections:

Quick start
The minimal properties needed for configuring a cluster.
Commonly used
Properties most frequently used when configuring Cassandra.
Performance tuning
Tuning performance and system resource utilization, including commit log, compaction, memory, disk I/O, CPU, reads, and writes.
Advanced
Properties for advanced users or properties that are less commonly used.
Security
Securing the database provides authentication, authorization for Cassandra.
Continuous paging Properties configure memory, threads, and duration when pushing pages continuously to the client.

Note: Values with ^note mark default values that are defined internally, missing, or commented out, or whose implementation depends on other properties in the cassandra.yaml file. Additionally, some commented-out values may not match the actual default values. These are recommended alternatives to the default values.

Quick start properties

The minimal properties needed for configuring a cluster.

cluster_name

Default: Test Cluster The name of the cluster. This setting prevents nodes in one logical cluster from joining another. All nodes in a cluster must have the same value.

listen_address

Default: localhost The IP address or hostname that the database binds to for connecting this node to other nodes.

Warning:

Never set listen_address to 0.0.0.0.
Set listen_address or listen_interface, do not set both.

Correct settings for various use cases:

Single-node installations: do one of the following:
- Comment this property out. If the node is properly configured (host name, name resolution, and so on), the database uses InetAddress.getLocalHost() to get the local address from the system.
- Leave set to the default, localhost.
Node in a multi-node installations: set this property to the node's IP address or hostname, or set listen_interface.
Node in a multi-network or multi-Datacenter installation, within an EC2 environment that supports automatic switching between public and private interfaces: set listen_address to the node's IP address or hostname, or set listen_interface.
Node with two physical network interfaces in a multi-datacenter installation or cluster deployed across multiple Amazon EC2 regions using the Ec2MultiRegionSnitch:
1. Set listen_address to this node's private IP or hostname, or set listen_interface (for communication within the local datacenter).
2. Set broadcast_address to the second IP or hostname (for communication between datacenters).
3. Set listen_on_broadcast_address to true.
4. If this node is a seed node, add the node's public IP address or hostname to the seeds list.
Open the storage_port or ssl_storage_port on the public IP firewall.

listen_interface

Default: eth0. ^note The interface that the database binds to for connecting to other nodes. Interfaces must correspond to a single address — IP aliasing is not supported.

Warning: Set listen_address or listen_interface, do not set both.

Default directories

If you have changed any of the default directories during installation, set these properties to the new locations. Make sure you have root access.

cdc_raw_directory

The directory where the CDC log is stored. Default locations:

The CDC file is located in the installation_location/data/cdc_raw directory.

The directory where Change Data Capture logs are stored.

commitlog_directory

The directory where the commit log is stored. Default location: /var/lib/cassandra/commitlog

For optimal write performance, place the commit log be on a separate disk partition, or (ideally) a separate physical device from the data file directories. Because the commit log is append only, an HDD is acceptable for this purpose.

data_file_directories

The directory location where table data is stored (in SSTables). The database distributes data evenly across the location, subject to the granularity of the configured compaction strategy. Default locations: /var/lib/cassandra/data.

For production, DataStax recommends RAID 0 and SSDs.

saved_caches_directory

The directory location where table key and row caches are stored.

Commonly used properties

Properties most frequently used when configuring Cassandra.

Before starting a node for the first time, you should carefully evaluate your requirements.

Common initialization properties

Note: Be sure to set the properties in the Quick start section as well.

disk_access_mode

Default: auto. Disk access mode.

Warning: Enabling mmap can cause excessive paging if all of the actively read SSTables do not fit into RAM.

auto
Enable mmap on both data and index files on a 64-bit JVM.
mmap
Map index and data files.
mmap_index_only
Map only index files. Use this setting if you observe high number of page faults or steals along with increased latencies.
standard
Disable mmap entirely.

commit_failure_policy

Default: stop. Policy for commit disk failures:

die
Shut down gossip and Thrift and kill the JVM, so the node can be replaced.
stop
Shut down gossip and Thrift, leaving the node effectively dead, available for inspection using JMX.
stop_commit
Shut down the commit log, letting writes collect but continuing to service reads.
ignore
Ignore fatal errors and let the batches fail.

prepared_statements_cache_size_mb

Default: auto, which is 1/256th of the heap or 10 MB, whichever is greater. Maximum size of the native protocol prepared statement cache.

Note: Specifying a value that is too large results in long running GCs and possibly out-of-memory errors. Keep the value at a small fraction of the heap.

Constantly re-preparing statements is a performance penalty.

thrift_prepared_statements_cache_size_mb

Default: auto, which is 1/256th of the heap or 10 MB, whichever is greater. Maximum size of the Thrift prepared statement cache. Leave empty if you do not use Thrift.

Note: Specifying a value that is too large results in long running GCs and possibly out-of-memory errors. Keep the value at a small fraction of the heap.

Constantly re-preparing statements is a performance penalty.

disk_optimization_strategy

Default: disabled. ^note The strategy for optimizing disk reads. Possible values: ssd or spinning.

disk_failure_policy

Default: stop. Sets how the database responds to disk failure. Recommend settings: stop or best_effort. Valid values:

die
Shut down gossip and Thrift and kill the JVM for any file system errors or single SSTable errors, so the node can be replaced.
stop_paranoid
Shut down gossip and Thrift even for single SSTable errors.
stop
Shut down gossip and Thrift, leaving the node effectively dead, but available for inspection using JMX.
best_effort
Stop using the failed disk and respond to requests based on the remaining available SSTables. This allows obsolete data at consistency level of ONE.
ignore
Ignore fatal errors and lets the requests fail; all file system errors are logged but otherwise ignored.

endpoint_snitch

Default: SimpleSnitch. Set to a class that implements the IEndpointSnitch interface. The database uses the snitch to locate nodes and route requests.

Important: Use only snitch implementations bundled with Cassandra.

SimpleSnitch
Only appropriate for Development deployments. Does not recognize datacenter or rack information.
GossipingPropertyFileSnitch
Recommended for production. Reads rack and datacenter for the local node in cassandra-rackdc.properties file and propagates these values to other nodes via gossip. For migration from the PropertyFileSnitch, uses the cassandra-topology.properties file if it is present.
PropertyFileSnitch
Determines proximity by rack and datacenter, which are explicitly configured in cassandra-topology.properties file.
Ec2Snitch
For EC2 deployments in a single region. Loads region and availability zone information from the Amazon EC2 API. The region is treated as the datacenter and the availability zone as the rack and uses only private IP addresses. For this reason, it does not work across multiple regions.
Ec2MultiRegionSnitch
Uses the public IP as the broadcast_address to allow cross-region connectivity. This means you must also set seed addresses to the public IP and open the storage_port or ssl_storage_port on the public IP firewall. For intra-region traffic, the database switches to the private IP after establishing a connection.
RackInferringSnitch:
Proximity is determined by rack and datacenter, which are assumed to correspond to the 3rd and 2nd octet of each node's IP address, respectively. Best used as an example for writing a custom snitch class (unless this happens to match your deployment conventions).
GoogleCloudSnitch:
Use for deployments on Google Cloud Platform across one or more regions. The region is treated as a datacenter and the availability zones are treated as racks within the datacenter. All communication occurs over private IP addresses within the same logical network.
CloudstackSnitch
Use the CloudstackSnitch for Apache Cloudstack environments.

Related information: Snitches

rpc_address

Default: localhost. The listen address for client connections (Thrift RPC service and native transport). Valid values:

unset:
Resolves the address using the configured hostname configuration of the node. If left unset, the hostname resolves to the IP address of this node using /etc/hostname, /etc/hosts, or DNS.
0.0.0.0:
Listens on all configured interfaces. You must set the broadcast_rpc_address to a value other than 0.0.0.0.
IP address
hostname

Related information: Network

rpc_interface

Default: eth1. ^note The listen address for client connections. Interface must correspond to a single address, IP aliasing is not supported. See rpc_address.

rpc_interface_prefer_ipv6

Default: false. Use IPv4 or IPv6 when interface is specified by name.

false - use first IPv4 address.
true - use first IPv6 address.

When only a single address is used, that address is selected without regard to this setting.

seed_provider

The addresses of hosts designated as contact points in the cluster. A joining node contacts one of the nodes in the -seeds list to learn the topology of the ring.

Important: Use only seed provider implementations bundled with Cassandra.

class_name (Default: org.apache.cassandra.locator.SimpleSeedProvider)
The class that handles the seed logic. It can be customized, but this is typically not required.
- seeds (Default: 127.0.0.1)
A comma-delimited list of IP addresses used by gossip for bootstrapping new nodes joining a cluster. If your cluster includes multiple nodes, you must change the list from the default value to the IP address of one of the nodes.
Attention: Making every node a seed node is not recommended because of increased maintenance and reduced gossip performance. Gossip optimization is not critical, but it is recommended to use a small seed list (approximately three nodes per datacenter).

enable_user_defined_functions

Default: false. User defined functions (UDFs) present a security risk, since they are executed on the server side. UDFs are executed in a sandbox to contain the execution of malicious code. They are disabled by default.

enable_scripted_user_defined_functions

Default: false. Java UDFs are always enabled, if enable_user_defined_functions is true. Enable this option to use UDFs with language javascript or any custom JSR-223 provider. This option has no effect if enable_user_defined_functions is false.

Common compaction settings

compaction_throughput_mb_per_sec: Default: 16. Throttles compaction to the specified Mb/second across the instance. The faster the database inserts data, the faster the system must compact in order to keep the SSTable count down. The recommended value is 16 to 32 times the rate of write throughput (in Mb/second). Setting the value to 0 disables compaction throttling.
Related information: Configuring compaction
compaction_large_partition_warning_threshold_mb: Default: 100. The database logs a warning when compacting partitions larger than the set value.

Common memtable settings

memtable_heap_space_in_mb: Default: 1/4 of heap size. ^note; The amount of on-heap memory allocated for memtables. The database uses the total of this amount and the value of memtable_offheap_space_in_mb to set a threshold for automatic memtable flush. For details, see memtable_cleanup_threshold.
Related information: Tuning the Java heap
memtable_offheap_space_in_mb: Default: 1/4 of heap size. ^note; Sets the total amount of off-heap memory allocated for memtables. The database uses the total of this amount and the value of memtable_heap_space_in_mb to set a threshold for automatic memtable flush. For details, see memtable_cleanup_threshold.
Related information: Tuning the Java heap

Common disk settings

concurrent_reads: Default: 32. ^note Workloads with more data than can fit in memory encounter a bottleneck in fetching data from disk during reads. Setting concurrent_reads to (16 × number_of_drives) allows operations to queue low enough in the stack so that the OS and drives can reorder them. The default setting applies to both logical volume managed (LVM) and RAID drives.
concurrent_writes: Default: 32. ^note Writes in Cassandra are rarely I/O bound, so the ideal number of concurrent writes depends on the number of CPU cores on the node. The recommended value is 8 × number_of_cpu_cores.
concurrent_counter_writes: Default: 32. ^note Counter writes read the current values before incrementing and writing them back. The recommended value is (16 × number_of_drives) .
concurrent_materialized_view_writes: Default: 32. Limit on the number of concurrent materialized view writes. Set this to the lesser of concurrent reads or concurrent writes, because there is a read involved in each materialized view write.

Common automatic backup settings

incremental_backups: Default: false. Backs up data updated since the last snapshot was taken. When enabled, the database creates a hard link to each SSTable flushed or streamed locally in a backups subdirectory of the keyspace data. Removing these links is the operator's responsibility.
Related information: Enabling incremental backups
snapshot_before_compaction: Default: false. Enables or disables taking a snapshot before each compaction. A snapshot is useful to back up data when there is a data format change. Be careful using this option: the database does not clean up older snapshots automatically.
Related information: Configuring compaction
snapshot_before_dropping_column: When enabled, every time the user drops a column/columns from a table, a snapshot is created on each node in the cluster before the change in schema is applied. Those snapshots have the same name on each node. For example: auto-snapshot_drop-column-columnname_20200515143511000. The name includes the name of the dropped column and the timestamp (UTC) when the column was dropped.
Important: The database does not automatically clear incremental backup files. DataStax recommends setting up a process to clear incremental backup hard links each time a new snapshot is created.
Default: false

Common fault detection setting

phi_convict_threshold: Default: 8. ^note Adjusts the sensitivity of the failure detector on an exponential scale. Generally this setting does not need adjusting.
Related information: About failure detection and recovery

Performance tuning properties

Tuning performance and system resource utilization, including commit log, compaction, memory, disk I/O, CPU, reads, and writes.

Commit log settings

commitlog_sync

Commit log synchronization method:

periodic - Send ACK signal for writes immediately. Commit log is synced every commitlog_sync_period_in_ms.
batch - Send ACK signal for writes after the commit log has been flushed to disk. Each incoming write triggers the flush task.

Default: periodic

commitlog_sync_period_in_ms

Default: 10000. Use with

commitlog_sync:
                                                  periodic

. Controls how often the commit log is synchronized to disk. Periodic syncs are acknowledged immediately.

commitlog_sync_batch_window_in_ms

Default: disabled. Use with commitlog_sync:batch. The maximum length of time that queries may be batched together.

commitlog_segment_size_in_mb

Disk usage threshold for commit logs before triggering the database flushing memtables to disk. If the total space used by all commit logs exceeds this threshold, the database flushes memtables to disk for the oldest commitlog segments to reclaim disk space by removing those log segments from the commit log. This flushing reduces the amount of data to replay on start-up, and prevents infrequently updated tables from keeping commitlog segments indefinitely.

Tip: See Configuring memtable thresholds and Commit log archive configuration.

The size of an individual commitlog file segment. A commitlog segment may be archived, deleted, or recycled after all its data has been flushed to SSTables. This data can potentially include commitlog segments from every table in the system. The default size is usually suitable for most commitlog archiving, but if you want a finer granularity, 8 or 16 MB is reasonable.

Default: 32

max_mutation_size_in_kb

Default: ½ of commitlog_segment_size_in_mb.

If a mutation's size exceeds this value, the mutation is rejected. Before increasing the commitlog segment size of the commitlog segments, investigate why the mutations are larger than expected. Look for underlying issues with access patterns and data model, because increasing the commitlog segment size is a limited fix.

Restriction: If you set max_mutation_size_in_kb explicitly, then you must set commitlog_segment_size_in_mb to at least twice the size of max_mutation_size_in_kb / 1024.

See commitlog_segment_size_in_mb above.

commitlog_compression

The compressor to use if commit log is compressed. To make changes, uncomment the commitlog_compression section and make changes to these options:

# commitlog_compression:
#   - class_name: LZ4Compressor
#     parameters:
#         -

class_name: LZ4Compressor, Snappy, or Deflate
parameters: optional parameters for the compressor

When not set, the default compression for the commit log is uncompressed.

Default: commented out

Lightweight transactions (LWT) settings

# concurrent_lw_transactions: 128
# max_pending_lw_transactions: 10000

concurrent_lw_transactions

Maximum number of permitted concurrent lightweight transactions (LWT).

A higher number might improve throughput if non-contending LWTs are in heavy use, but will use more memory and might be less successful with contention.
When not set, the default value is 8x the number of TPC cores. This default value is appropriate for most environments.

Default: calculated 8x the number of TPC cores

max_pending_lw_transactions

Maximum number of lightweight transactions (LWT) in the queue before node reports OverloadedException for LWTs.

Default: 10000

Change-data-capture (CDC) space settings

cdc_enabled: false
cdc_total_space_in_mb: 4096
cdc_free_space_check_interval_ms: 250

Advanced properties

Properties for advanced users or properties that are less commonly used.

Advanced initialization properties

auto_bootstrap

Default: true. This setting has been removed from default configuration. It causes new (non-seed) nodes migrate the right data to themselves automatically. When initializing a fresh cluster without data, add auto_bootstrap: false.

batch_size_warn_threshold_in_kb

Default: 64. Log a warning message when any multiple-partition batch size exceeds this value.

CAUTION: Increasing this threshold can lead to node instability.

batch_size_fail_threshold_in_kb

Default: 640. Fails any batch whose size exceeds this setting. The default value is 10X the value of batch_size_warn_threshold_in_kb.

unlogged_batch_across_partitions_warn_threshold

Default: 10. Causes the database to log a WARN message on any batches not of type LOGGED that span across more partitions than this limit. The default value is 10 partitions.

cdc_enabled

Default: false. Enable or disable change data capture (CDC) functionality on a per-node basis. This modifies the logic used for write path allocation rejection (standard: never reject. cdc: reject Mutation containing a CDC-enabled table if at space limit in cdc_raw_directory).

Important: Do not enable CDC on a mixed-version cluster. Upgrade all nodes to Cassandra before enabling and restarting the cluster.

broadcast_address

Default: listen_address. ^note The public IP address this node uses to broadcast to other nodes outside the network or across regions in multiple-region EC2 deployments. If this property is commented out, the node uses the same IP address or hostname as listen_address. A node does not need a separate broadcast_address in a single-node or single-datacenter installation, or in an EC2-based network that supports automatic switching between private and public communication. It is necessary to set a separate listen_address and broadcast_address on a node with multiple physical network interfaces or other topologies where not all nodes have access to other nodes by their private IP addresses. For specific configurations, see the instructions for listen_address.

listen_on_broadcast_address

Default: false. If this node uses multiple physical network interfaces, set a unique IP address for broadcast_address and set listen_on_broadcast_address to true. This enables the node to communicate on both interfaces.

Set this property to false if the node is on a network that automatically routes between public and private networks, as Amazon EC2 does.

For configuration details, see the instructions for listen_address.

initial_token

Default: 1 (disabled). Set this property for single-node-per-token architecture, in which a node owns exactly one contiguous range in the ring space. Setting this property overrides num_tokens.

If your installation is not using vnodes or this node's num_tokens is set it to 1 or is commented out, you should always set an initial_token value when setting up a production cluster for the first time, and when adding capacity. See Generating tokens.

Use this parameter only with num_tokens (vnodes) in special cases such as Restoring from a snapshot.

num_tokens

Default: 1 (disabled)

Define virtual node (vnode) token architecture.

Note: All other nodes in the datacenter must have the same token architecture.

1 - disable vnodes and use 1 token for legacy compatibility.
a number between 2 and 128 - the number of token ranges to assign to this virtual node (vnode). A higher value increases the probability that the data and workload are evenly distributed.
CAUTION: Using vnodes can impact performance for your cluster. DataStax recommends testing the configuration before enabling vnodes in production environments.
When the token number varies between nodes in a datacenter, the vnode logic assigns a proportional number of ranges relative to other nodes in the datacenter. In general, if all nodes have equal hardware capability, each node should have the same num_tokens value.

To migrate an existing cluster from single node per token range to vnodes, see Enabling virtual nodes on an existing production cluster.

allocate_tokens_for_local_replication_factor

RF of keyspaces in datacenter - triggers the recommended algorithmic allocation for the RF and num_tokens for this node.
The allocation algorithm optimizes the workload balance using the target keyspace replication factor. DataStax recommends setting the number of tokens to 8 to distribute the workload with ~10% variance between nodes. The allocation algorithm attempts to choose tokens in a way that optimizes replicated load over the nodes in the datacenter for the specified RF. The load assigned to each node is close to proportional to the number of vnodes.
Note: The allocation algorithm is supported only for the Murmur3Partitioner and RandomPartitioner partitioners. The Murmur3Partitioner is the default partitioning strategy for new Cassandra clusters and the right choice for new clusters in almost all cases.
commented out - uses the random selection algorithm to assign token ranges randomly.
Note: Over time, loads in a datacenter using the random selection algorithm become unevenly distributed. DataStax recommends using only the allocation algorithm.

See Virtual node (vnode) configuration, and for set up instructions see Adding vnodes to an existing cluster or Adding a datacenter to a cluster.

partitioner

Default: org.apache.cassandra.dht.Murmur3Partitioner. Sets the class that distributes rows (by partition key) across all nodes in the cluster. Any IPartitioner may be used, including your own as long as it is in the class path. For new clusters use the default partitioner.

Cassandra provides the following partitioners for backwards compatibility:

RandomPartitioner
ByteOrderedPartitioner (deprecated)
OrderPreservingPartitioner (deprecated)

Important: Use only partitioner implementations bundled with Cassandra.

Related information: Partitioners

storage_port

Default: 7000. The port for inter-node communication.

tracetype_query_ttl

Default: 86400. TTL for different trace types used during logging of the query process.

tracetype_repair_ttl

Default: 604800. TTL for different trace types used during logging of the repair process.

Advanced automatic backup setting

auto_snapshot: Default: true. Enables or disables whether the database takes a snapshot of the data before truncating a keyspace or dropping a table. To prevent data loss, DataStax strongly advises using the default setting. If you set auto_snapshot to false, you lose data on truncation or drop.

Key caches and global row properties

When creating or modifying tables, you can enable or disable the key cache (partition key cache) or row cache for that table by setting the caching parameter. Other row and key cache tuning and configuration options are set at the global (node) level. The database uses these settings to automatically distribute memory for each table on the node based on the overall workload and specific table usage. You can also configure the save periods for these caches globally.

Related information: Configuring caches

key_cache_keys_to_save: Default: disabled. All keys are saved. ^note Number of keys from the key cache to save.
key_cache_save_period: Default: 14400. (4 hours) Duration in seconds that keys are kept in cache. Caches are saved to saved_caches_directory. Saved caches greatly improve cold-start speeds and have relatively little effect on I/O.
key_cache_size_in_mb: Default: empty. A global cache setting for the maximum size of the key cache in memory (for all tables). If no value is set, the cache is set to the smaller of 5% of the available heap, or 100MB. To disable set to 0.
Related information: nodetool setcachecapacity, Enabling and configuring caching.
column_index_cache_size_in_kb: Default: 2. A threshold for the total size of all index entries for a partition that the database stores in the partition key cache. If the total size of all index entries for a partition exceeds this amount, the database stops putting entries for this partition into the partition key cache. This limitation prevents index entries from large partitions from taking up all the space in the partition key cache (which is controlled by key_cache_size_in_mb).
row_cache_class_name: Default: disabled. ^note The classname of the row cache provider to use. Valid values: OHCProvider (fully off-heap) or SerializingCacheProvider (partially off-heap).
Important: Use only row cache provider implementations bundled with Cassandra.
row_cache_keys_to_save: Default: disabled. All keys are saved. ^note Number of keys from the row cache to save.
row_cache_size_in_mb: Default: 0. To disable, set to 0. Maximum size of the row cache in memory. The row cache can save more time than key_cache_size_in_mb,, but it is space-intensive because it contains the entire row. Use the row cache only for hot rows or static rows. If you reduce the size, you may not get you hottest keys loaded on start up.
row_cache_save_period: Default: 0. To disable, set to 0. The number of seconds that rows are kept in cache. Caches are saved to saved_caches_directory. This setting has limited use as described in row_cache_size_in_mb.

Counter caches properties

Counter cache helps to reduce counter locks' contention for hot counter cells. In case of RF = 1 a counter cache hit causes the database to skip the read before write entirely. With RF > 1 a counter cache hit still helps to reduce the duration of the lock hold, helping with hot counter cell updates, but does not allow skipping the read entirely. Only the local (clock, count) tuple of a counter cell is kept in memory, not the whole counter, so it is relatively cheap.

Note: If you reduce the counter cache size, the database may load the hottest keys start-up.

counter_cache_size_in_mb: Default value: empty. ^note When no value is set, the database uses the smaller of minimum of 2.5% of Heap or 50 megabytes (MB). If your system performs counter deletes and relies on low gc_grace_seconds, you should disable the counter cache. To disable, set to 0.
counter_cache_save_period: Default: 7200. (2 hours) the amount of time after which the database saves the counter cache (keys only). The database saves caches to saved_caches_directory.
counter_cache_keys_to_save: Default value: disabled. ^note Number of keys from the counter cache to save. When this property is disabled, the database saves all keys.

Tombstone settings

When executing a scan, within or across a partition, the database must keep tombstones in memory to allow them to return to the coordinator. The coordinator uses tombstones to ensure that other replicas know about the deleted rows. Workloads that generate numerous tombstones may cause performance problems and exhaust the server heap. See Cassandra anti-patterns: Queues and queue-like datasets. Adjust these thresholds only if you understand the impact and want to scan more tombstones. You can adjust these thresholds at runtime using the StorageServiceMBean.

tombstone_warn_threshold: Default: 1000. The database issues a warning if a query scans more than this number of tombstones.
tombstone_failure_threshold: Default: 100000. The database aborts a query if it scans more than this number of tombstones.

Network timeout settings

aggregated_request_timeout_in_ms: Number of milliseconds that the coordinator waits for aggregated read operations to complete. For example, SELECT COUNT(*), MIN(x). Default: 120000.
cas_contention_timeout_in_ms: Default: 1000. The number of milliseconds during which the coordinator continues to retry a CAS (compare and set) operation that contends with other proposals for the same row. If the coordinator cannot complete the operation within this timespan, it aborts the operation.
counter_write_request_timeout_in_ms: Default: 5000. The number of milliseconds that the coordinator waits for counter writes to complete before timing it out.
range_request_timeout_in_ms: Default: 10000. The number of milliseconds that the coordinator waits for sequential or index scans to complete before timing it out.
read_request_timeout_in_ms: Default: 5000. The number of milliseconds that the coordinator waits for read operations to complete before timing it out.
request_timeout_in_ms: Default: 10000. The default timeout value for other miscellaneous operations.
Related information: Hinted handoff: repair during write path.
slow_query_log_timeout_in_ms: Default: 500. How long before a node logs slow queries. Select queries that exceed this value generate an aggregated log message to identify slow queries. To disable, set to 0.
truncate_request_timeout_in_ms: Default: 60000. The number of milliseconds that the coordinator waits for a truncate (the removal of all data from a table) to complete before timing it out. The long default value allows the database to take a snapshot before removing the data. If auto_snapshot is disabled (not recommended), you can reduce this time.
write_request_timeout_in_ms: Default: 2000. The number of milliseconds that the coordinator waits for a write operations to complete before timing it out for requests with at least one node in the local datacenter.
Related information: Hinted handoff: repair during write path.
cross_dc_rtt_in_ms: Default: 0. Increases the cross-datacenter timeout (write_request_timeout_in_ms + cross_dc_rtt_in_ms) for requests that only involve nodes in a remote datacenter. This setting is intended to reduce hint pressure.
Tip: DataStax recommends using LOCAL_* consistency levels (CL) for read and write requests in multi-datacenter deployments to avoid timeouts that may occur when remote nodes are chosen to satisfy the CL, such as QUORUM.

Inter-node settings

cross_node_timeout

Default: false. Enables or disables operation timeout information exchange between nodes (to accurately measure request timeouts). If this property is disabled, the replica assumes any requests are forwarded to it instantly by the coordinator. During overload conditions this means extra time is required for processing already-timed-out requests..

CAUTION: Before enabling this property make sure NTP (network time protocol) is installed and the times are synchronized among the nodes.

internode_send_buff_size_in_bytes

Default: empty. ^note The sending socket buffer size in bytes for inter-node calls.

The buffer size set by this parameter and internode_recv_buff_size_in_bytes is limited by net.core.wmem_max. If this property is not set, net.ipv4.tcp_wmem determines the buffer size. For more details run

man
                                                  tcp

and refer to:

/proc/sys/net/core/wmem_max
/proc/sys/net/core/rmem_max
/proc/sys/net/ipv4/tcp_wmem
/proc/sys/net/ipv4/tcp_wmem

Related information: TCP settings

internode_recv_buff_size_in_bytes

Default: empty. ^note The receiving socket buffer size in bytes for inter-node calls.

internode_compression

Default: dc. Controls whether traffic between nodes is compressed. Valid values:

all
Compresses all traffic.
dc
Compresses traffic between datacenters only.
none
No compression.

inter_dc_tcp_nodelay

Default: false. Enable this property or disable tcp_nodelay for inter-datacenter communication. If this property is disabled, the network sends larger, but fewer, network packets. This reduces overhead from the TCP protocol itself. However, disabling inter_dc_tcp_nodelay may increase latency by blocking cross datacenter responses.

Native transport (CQL Binary Protocol)

start_native_transport

Default: true. Enables or disables the native transport server. This server uses the same address as the rpc_address, but the port it uses is different from rpc_port. See native_transport_port.

native_transport_port

Default: 9042. The port where the CQL native transport listens for clients.

native_transport_max_threads

Default: 128. ^note The maximum number of thread handling requests. Similar to rpc_max_threads, but this property differs as follows:

The default for native_transport_max_threads is 128; the default for rpc_max_threads is unlimited.
There is no corresponding native_transport_min_threads.
The database stops idle native transport threads after 30 seconds.

native_transport_max_frame_size_in_mb

Default: 256. The maximum allowed size of a frame. Frame (requests) larger than this are rejected as invalid.

native_transport_max_concurrent_connections

Default: -1. The maximum number of concurrent client connections. The default value of -1 means unlimited.

native_transport_max_concurrent_connections_per_ip

Default: -1. The maximum number of concurrent client connections per source IP address. The default value of -1 means unlimited.

RPC (remote procedure call) settings

Settings for configuring and tuning client connections.

broadcast_rpc_address

Default: empty. ^note The RPC address for broadcast to drivers and other nodes. This cannot be set to 0.0.0.0. If left blank, the database uses the rpc_address or rpc_interface. If rpc_address or rpc_interface is set to 0.0.0.0, this property must be set.

rpc_port

Default: 9160. Thrift port for client connections.

start_rpc

Default: true. Enables or disables the Thrift RPC server.

rpc_keepalive

Default: true. Enables or disables keepalive on client connections (RPC or native).

rpc_max_threads

Default: unlimited. ^note Regardless of your choice of RPC server (rpc_server_type), rpc_max_threads dictates the maximum number of concurrent requests in the RPC thread pool. If you are using the parameter sync (see rpc_server_type) it also dictates the number of clients that can be connected. A high number of client connections could cause excessive memory usage for the thread stack. Connection pooling on the client side is highly recommended. Setting a rpc_max_threads acts as a safeguard against misbehaving clients. If the number of threads reaches the maximum, the database blocks additional connections until a client disconnects.

rpc_min_threads

Default: unlimited. ^note The minimum thread pool size for remote procedure calls.

rpc_recv_buff_size_in_bytes

Default: empty. ^note The receiving socket buffer size for remote procedure calls.

rpc_send_buff_size_in_bytes

Default: empty. ^note The sending socket buffer size in bytes for remote procedure calls.

rpc_server_type

Default: sync. The database provides three options for the RPC server. sync and hsha performance is about the same, but hsha uses less memory.

sync: (Default: one thread per Thrift connection.)
For a very large number of clients, memory is the limiting factor. On a 64-bit JVM, 180 KB is the minimum stack size per thread and corresponds to your use of virtual memory. Physical memory may be limited depending on use of stack space.
hsha:
Half synchronous, half asynchronous. All Thrift clients are handled asynchronously using a small number of threads that does not vary with the number of clients. This mechanism scales well to many clients. The RPC requests are synchronous (one thread per active request).
Note: If you select this option, you must change the default value (unlimited) of rpc_max_threads.
Your own RPC server
You must provide a fully-qualified class name of an o.a.c.t.TServerFactory that can create a server instance.

Advanced fault detection settings

Settings to handle poorly performing or failing components.

gc_warn_threshold_in_ms

Default: 1000. Any GC pause longer than this interval is logged at the WARN level. (By default, the database logs any GC pause greater than 200 ms at the INFO level.)

Additional information: Configuring logging.

otc_coalescing_strategy

Default: DISABLED. Supported strategies are: FIXED, MOVINGAVERAGE, TIMEHORIZON, and DISABLED. Suitable for VMs, but not noticeably performant in other environments. The OutboundTcpConnection (otc) strategy to:

Increase message throughput (doubling or more).
Process multiple messages with one trip to read from a socket.
Perform all the task submission work at the same time.
Reduce context switching.
Increase cache friendliness of network message processing.

Important: Use only strategy implementations bundled with Cassandra.

otc_coalescing_window_us

Default: disabled. ^note How many microseconds to wait for coalescing. For fixed strategy, the amount of time after the first message is received before it is sent with any accompanying messages. For moving average, this is the maximum wait time and the interval that messages must arrive on average to enable coalescing.

otc_coalescing_enough_coalesced_messages

Default: disabled. ^note The threshold for the number of messages. Do not coalesce messages when this value is exceeded. Should be more than 2 and less than 128.

seed_gossip_probability

The percentage of time that gossip messages are sent to a seed node during each round of gossip. Decreases the time to propagate gossip changes across the cluster. Default: 1.0 (100%)

back_pressure_enabled

Default: false. Enable for the coordinator to apply the specified back pressure strategy to each mutation that is sent to replicas.

back_pressure_strategy

back_pressure_strategy:
    - class_name: org.apache.cassandra.net.RateBasedBackPressure
      parameters:
        - high_ratio: 0.90
          factor: 5
          flow: FAST

Default: RateBasedBackPressure. To add new strategies, implement org.apache.cassandra.net.BackpressureStrategy and provide a public constructor that accepts a Map<String, Object>.

Important: Use only strategy implementations bundled with Cassandra.

RateBasedBackPressure
Ratio between incoming mutation responses and outgoing mutation requests.
high_ratio
When outgoing mutations are below this value, they are rate limited according to the incoming rate decreased by the factor (described below). When above this value, the rate limiting is increased by the factor.
factor
A number between 1 and 10. Increases or decreases rate limiting.
flow
Default: FAST. The flow speed to apply rate limiting:
- FAST - rate limited to the speed of the fastest replica.
- SLOW - rate limit to the speed of the slowest replica.

max_value_size_in_mb

Default: 256. ^note The maximum size of any value in SSTables. It detects SSTable corruption and marks the SSTables as corrupted when the threshold is exceeded.

dynamic_snitch_badness_threshold

Default: 0.1. The performance threshold for dynamically routing client requests away from a poorly performing node. Specifically, it controls how much worse a poorly performing node has to be before the dynamic snitch prefers other replicas. A value of 0.2 means the database continues to prefer the static snitch values until the node response time is 20% worse than the best performing node. Until the threshold is reached, incoming requests are statically routed to the closest replica as determined by the snitch. A value of zero to 1.0 for the read_repair_chance table property maximizes cache capacity across the nodes.

dynamic_snitch_reset_interval_in_ms

Default: 600000. Time interval after which the database resets all node scores. This allows a bad node to recover.

dynamic_snitch_update_interval_in_ms

Default: 100. The number of milliseconds between the database's calculation of node scores. Because score calculation is CPU intensive, be careful when reducing this interval.

hints_flush_period_in_ms

Default: 10000. The number of milliseconds the database waits before flushing hints from internal buffers to disk.

hints_directory

The directory in which hints are stored. If not set, the directory is $CASSANDRA_HOME/data/hints.

Default: /var/lib/cassandra/hints

hinted_handoff_enabled

Enables or disables hinted handoff. A hint indicates that the write needs to be replayed to an unavailable node. The database writes the hint to a hints file on the coordinator node.

false - do not enable hinted handoff
true - globally enable hinted handoff, except for datacenters specified for hinted_handoff_disabled_datacenters

Default: true

hinted_handoff_disabled_datacenters

A blacklist of datacenters that will not perform hinted handoffs. To disable hinted handoff on a certain datacenter, add its name to this list.

Default: commented out

hinted_handoff_throttle_in_kb: Default: 1024. Maximum amount of traffic per delivery thread in kilobytes per second. This rate reduces proportionally to the number of nodes in the cluster. For example, if there are two nodes in the cluster, each delivery thread uses the maximum rate. If there are three, each node throttles to half of the maximum, since the two nodes are expected to deliver hints simultaneously.
Note: When applying this limit, the database computes the hint transmission rate based on the uncompressed hint size, even if internode_compression or hints_compression is enabled.
max_hint_window_in_ms: Default: 10800000. (3 hours) Maximum amount of time during which the database generates hints for an unresponsive node. After this interval, the database does not generate any new hints for the node until it is back up and responsive. If the node goes down again, the database starts a new interval. This setting can prevent a sudden demand for resources when a node is brought back online and the rest of the cluster attempts to replay a large volume of hinted writes.
Related information: About failure detection and recovery
max_hints_delivery_threads: Default: 2. Number of threads the database uses to deliver hints. In multiple datacenter deployments, consider increasing this number because cross datacenter handoff is generally slower.
max_hints_file_size_in_mb: Default: 128. The maximum size for a single hints file, in megabytes.
hints_compression: Default: LZ4Compressor. The compressor for hint files. Supported compressors: LZ, Snappy, and Deflate. If you do not specify a compressor, the database does not compress hints files.
batchlog_replay_throttle_in_kb: (Default: 1024 kilobytes per second) Total maximum throttle for replaying failed logged batches. Throttling is reduced proportionally to the number of nodes in the cluster.

Request scheduler properties

Settings to handle incoming client requests according to a defined policy. If your nodes are overloaded and dropping requests, DataStax recommends that you add more nodes rather than use these properties to prioritize requests.

Note: The properties in this section apply only to the Thrift transport. They have no effect on the use of CQL over the native protocol.

request_scheduler

Default: org.apache.cassandra.scheduler.NoScheduler. The scheduler to handle incoming client requests according to a defined policy. This scheduler is useful for throttling client requests in single clusters containing multiple keyspaces. This parameter is specifically for requests from the client and does not affect inter-node communication. Valid values:

org.apache.cassandra.scheduler.NoScheduler
The database does no scheduling.
org.apache.cassandra.scheduler.RoundRobinScheduler
The database uses a round robin of client requests to a node with a separate queue for each request_scheduler_id property.
The database uses a Java class that implements the RequestScheduler interface.

request_scheduler_id

Default: keyspace. ^note The scope of the scheduler's activity. Currently the only valid value is keyspace.

request_scheduler_options

Default: NoScheduler. A list of properties that define configuration options for request_scheduler.

RoundRobin:

A round robin of client requests to a node with a separate queue for each request_scheduler_id.

throttle_limit - The number of in-flight requests per client. Requests that exceed this limit are queued up until running requests complete. Recommended value is ((concurrent_reads + concurrent_writes) × 2).
default_weight - Default: 1. ^note How many requests the scheduler handles during each turn of the round robin.
weights - A list of keyspaces with assigned weights.

Thrift interface properties

Legacy API for older clients. CQL is a simpler and better API for the database.

thrift_framed_transport_size_in_mb: Default: 15. Frame size (maximum field length) for Thrift. The frame is the row or part of the row that the application is inserting.

Security properties

These properties protect Cassandra databases against potential harm due to deliberate attack or user error.

authenticator

(Default: AllowAllAuthenticator) The authentication backend. It implements IAuthenticatorfor identifying users. Available authenticators:

AllowAllAuthenticator. Disables authentication; Cassandra performs no checks.
PasswordAuthenticator. Authenticates users with user names and hashed passwords stored in thesystem_auth.credentials table. Leaving the default replication factor of 1 set for the system_auth keyspace results in denial of access to the cluster if the single replica of the keyspace goes down. For multiple datacenters, be sure to set the replication class toNetworkTopologyStrategy.

internode_authenticator

Default: enabled. ^note Internode authentication backend. It implements org.apache.cassandra.auth.AllowAllInternodeAuthenticator to allows or disallow connections from peer nodes.

authorizer

(Default: AllowAllAuthorizer) The authorization backend. It implements IAuthenticator to limit access and provide permissions. Available authorizers:

AllowAllAuthorizer Disables authorization. Cassandra allows any action to any user.
CassandraAuthorizer. Stores permissions in system_auth.permissions table. Leaving the default replication factor of 1 set for the system_auth keyspace results in denial of access to the cluster if the single replica of the keyspace goes down. For multiple datacenters, be sure to set the replication class to NetworkTopologyStrategy.

role_manager

(Default: CassandraRoleManager) Part of the Authentication & Authorization backend that implements IRoleManager to maintain grants and memberships between roles. Out of the box, Cassandra provides org.apache.cassandra.auth.CassandraRoleManager, which stores role information in the system_auth keyspace. Most functions of the IRoleManager require an authenticated login, so unless the configured IAuthenticator actually implements authentication, most of this functionality will be unavailable. CassandraRoleManager stores role data in the system_auth keyspace. If you use the role manager, increase system_auth keyspace replication factor .

roles_update_interval_in_ms

Default: 2000. Enable to refresh interval for roles cache. Defaults to the same value as roles_validity_in_ms. After this interval, cache entries become eligible for refresh. On next access, the database schedules an async reload, and returns the old value until the reload completes. If roles_validity_in_ms is non-zero, then this must be also.

credentials_validity_in_ms

Default: 2000. How many milliseconds credentials in the cache remain valid. This cache is tightly coupled to the provided PasswordAuthenticator implementation of IAuthenticator. If another IAuthenticator implementation is configured, the database does not use this cache, and these settings have no effect.

Note: Credentials are cached in encrypted form. This may cause a performance penalty that offsets the reduction in latency gained by caching.

CAUTION: Cache credentials and permissions are not automatically invalidated after issuing a REVOKE statement.

This setting is disabled when set to 0.

credentials_update_interval_in_ms

Default: same value as credentials_validity_in_ms. After this interval, cache entries become eligible for refresh. The next time the cache is accessed, the system schedules an asynchronous reload of the cache. Until this cache reload is complete, the cache returns the old values.

If credentials_validity_in_ms is nonzero, this property must also be nonzero.

permissions_validity_in_ms

Default: 2000. Fetching permissions can be resource intensive. Define how many milliseconds permissions in cache remain valid to manage performance impact of permissions queries. Set the cache validity period to your security tolerances. The cache is used for the standard authentication and the row-level access control (RLAC) cache. The cache is quite effective at small durations.

CAUTION: Cache credentials and permissions are not automatically invalidated after issuing a REVOKE statement.

This setting is disabled when set to 0.

permissions_update_interval_in_ms

Default: same value as permissions_validity_in_ms. Sets refresh interval for the standard authentication cache and the row-level access control (RLAC) cache. After this interval, cache entries become eligible for refresh. On next access, the database schedules an async reload and returns the old value until the reload completes. If permissions_validity_in_ms is nonzero, roles_update_interval_in_ms must also be non-zero.

server_encryption_options

Configure inter-node encryption. If enabled, you must also generate keys and provide the appropriate key and truststore locations and passwords. No custom encryption options are supported. Available options:

internode_encryption: Default: none. Enables or disables encryption of inter-node communication using the TLS_RSA_WITH_AES_128_CBC_SHA cipher suite for authentication, key exchange, and encryption of data transfers. Use the DHE/ECDHE ciphers, such as TLS_DHE_RSA_WITH_AES_128_CBC_SHA if running in (Federal Information Processing Standard) FIPS 140 compliant mode. Available inter-node options:
- all
  Encrypt all inter-node communications.
- none
  No encryption.
- dc
  Encrypt the traffic between the datacenters (server only).
- rack
  Encrypt the traffic between the racks (server only).
keystore: Default: conf/.keystore.
The location of a Java keystore (JKS) suitable for use with Java Secure Socket Extension (JSSE), which is the Java version of the Secure Sockets Layer (SSL), and Transport Layer Security (TLS) protocols. The keystore contains the private key used to encrypt outgoing messages.
keystore_password: Default: cassandra.
Password for the keystore.
truststore: Default: conf/.truststore.
Location of the truststore containing the trusted certificate for authenticating remote servers.
truststore_password: Default: cassandra.
Password for the truststore.

The passwords used in these options must match the passwords used when generating the keystore and truststore. For instructions on generating these files, see Creating a Keystore to Use with JSSE.

Advanced settings:

protocol: Default: TLS.
algorithm: Default: SunX509.
store_type: Default: JKS.
cipher_suites: Supported ciphers:
- TLS_RSA_WITH_AES_128_CBC_SHA
- TLS_RSA_WITH_AES_256_CBC_SHA
- TLS_DHE_RSA_WITH_AES_128_CBC_SHA
- TLS_DHE_RSA_WITH_AES_256_CBC_SHA
- TLS_ECDHE_RSA_WITH_AES_128_CBC_SHA
- TLS_ECDHE_RSA_WITH_AES_256_CBC_SHA
require_client_auth: Default: false.
Enables or disables certificate authentication.
require_endpoint_verification: Default: false.
Enables or disables host name verification.

Related information: Node-to-node encryption

client_encryption_options

Enables or disables client-to-node encryption. You must also generate keys and provide the appropriate key and truststore locations and passwords. Cassandra options:

enabled: Default: false.
To enable client encryption, set to true.
optional: Default: false.
When optional is selected, both encrypted and unencrypted connections over native transport are allowed. This is a necessary transition state to facilitate enabling client to node encryption on live clusters without inducing an outage for existing unencrypted clients. Typically, once existing clients are migrated to encrypted connections, optional is unselected in order to enforce native transport encryption.
keystore: Default: conf/.keystore.
The location of a Java keystore (JKS) suitable for use with Java Secure Socket Extension (JSSE), which is the Java version of the Secure Sockets Layer (SSL), and Transport Layer Security (TLS) protocols. The keystore contains the private key used to encrypt outgoing messages.
keystore_password: Default: cassandra.
Password for the keystore. This must match the password used when generating the keystore and truststore.
require_client_auth: Default: false.
Enables or disables certificate authentication.
truststore: Default: conf/.truststore.
Set this property if require_client_auth is true.
truststore_password: Default: cassandra
Set if require_client_auth is true.

Advanced settings:

protocol: Default: TLS.
algorithm: Default: SunX509.
store_type: Default: JKS.
cipher_suites: Supported ciphers:
- TLS_RSA_WITH_AES_128_CBC_SHA
- TLS_RSA_WITH_AES_256_CBC_SHA
- TLS_DHE_RSA_WITH_AES_128_CBC_SHA
- TLS_DHE_RSA_WITH_AES_256_CBC_SHA
- TLS_ECDHE_RSA_WITH_AES_128_CBC_SHA
- TLS_ECDHE_RSA_WITH_AES_256_CBC_SHA

Related information: Client-to-node encryption

transparent_data_encryption_options

Enables encryption of data at rest (on-disk). Recommendation: download and install the Java Cryptography Extension (JCE) Unlimited Strength Jurisdiction Policy Files for your version of the JDK.

TDE properties:

enabled: (Default: false)
chunk_length_kb: (Default: 64)
cipher: options:
- AES
- CBC
- PKCS5Padding
key_alias: testing:1
iv_length: 16
Note: iv_length is commented out in the default cassandra.yaml file. Uncomment only if cipher is set to AES. The value must be 16 (bytes).
key_provider:
- class_name: org.apache.cassandra.security.JKSKeyProvider
  parameters:
  - keystore: conf/.keystore
  - keystore_password: cassandra
  - store_type: JCEKS
  - key_password: cassandra

ssl_storage_port

Default: 7001. The SSL port for encrypted communication. Unused unless enabled in encryption_options.

native_transport_port_ssl

Default: 9142. If client encryption is enabled and native_transport_port_ssl is disabled, the native_transport_port (default: 9042) will encrypt all traffic. To use both unencrypted and encrypted traffic, enable native_transport_port_ssl.

Continuous paging

continuous_paging

Pushes pages continuously to the client when requested by the client, parameters control:

Maximum memory used. Default: 60 ⨉ 4 ⨉ 8 = 1920 MB (max_concurrent_sessions ⨉ max_session_pages ⨉ max_page_size_mb).
Maximum number of threads.
Maximum duration for local queries.

Guidance:

If the client is not reading from the socket, the producer thread is blocked after it has prepared max_session_pages, up to max_client_wait_time_ms.
Because memtables and SSTables are used by the continuous paging query, you can define the maximum period of time during which memtables cannot be flushed and compacted SSTables cannot be deleted.
Maximum period of time = max_client_wait_time_ms + max_local_query_time_ms.
Consider adjusting max_local_query_time_ms and max_client_wait_time_ms when high write workloads exist on tables that have continuous paging requests.
If fewer threads exist than sessions (max_threads < max_concurrent_sessions), a session cannot execute until another one is swapped out.
Distributed queries (CL > ONE or non-local data) are swapped out after every page, while local queries at CL = ONE are swapped out after max_local_query_time_ms.
If the client is slow in reading pages, try increasing the delay by adjusting max_client_wait_time_ms.

Parameters:

max_concurrent_sessions
Default: 60. The maximum number of concurrent sessions. Additional sessions are rejected with an unavailable error.
max_session_pages
Default: 4.The maximum number of pages that can be buffered for each session.
max_page_size_mb
Default: 8. The maximum size of a page, in MB. If an individual CQL row is larger than this value, the page can be bigger than this value.
max_client_wait_time_ms
Default: 20000. The maximum time for the server to wait for the client to read from the socket. If exceeded, the session is aborted and the client receives an error. Setting max_client_wait_time_ms to a value too low may result in client side errors.
max_local_query_time_ms
Default: 5000. The maximum time for a local continuous query to run. When exceeded, the session is swapped out and rescheduled. Swapping and rescheduling ensures the release of resources that prevent the memtables from flushing and ensures fairness when max_threads < max_concurrent_sessions.
max_threads
Default: 24. The number of threads dedicated to continuous paging sessions.