Cluster settings

The following settings are related to the OpenSearch cluster.

To learn more about static and dynamic settings, see Configuring OpenSearch.

Cluster-level routing and allocation settings

OpenSearch supports the following cluster-level routing and shard allocation settings:

• cluster.routing.allocation.enable (Dynamic, string): Enables or disables allocation for specific kinds of shards.

  Valid values are:

  • all – Allows shard allocation for all types of shards.
  • primaries – Allows shard allocation for primary shards only.
  • new_primaries – Allows shard allocation for primary shards for new indexes only.
  • none – No shard allocations are allowed for any indexes.

  Default is all.

• cluster.routing.allocation.node_concurrent_incoming_recoveries (Dynamic, integer): Configures how many concurrent incoming shard recoveries are allowed to occur on a node. Default is 2.

• cluster.routing.allocation.node_concurrent_outgoing_recoveries (Dynamic, integer): Configures how many concurrent outgoing shard recoveries are allowed to occur on a node. Default is 2.

• cluster.routing.allocation.node_concurrent_recoveries (Dynamic, string): Used to set cluster.routing.allocation.node_concurrent_incoming_recoveries and cluster.routing.allocation.node_concurrent_outgoing_recoveries to the same value.

• cluster.routing.allocation.node_initial_primaries_recoveries (Dynamic, integer): Sets the number of recoveries for unassigned primaries after a node restart. Default is 4.

• cluster.routing.allocation.same_shard.host (Dynamic, Boolean): When set to true, multiple copies of a shard are prevented from being allocated to distinct nodes on the same host. Default is false.

• cluster.routing.rebalance.enable (Dynamic, string): Enables or disables rebalancing for specific kinds of shards.

  Valid values are:

  • all – Allows shard balancing for all types of shards.
  • primaries – Allows shard balancing for primary shards only.
  • replicas – Allows shard balancing for replica shards only.
  • none – No shard balancing is allowed for any indexes.

  Default is all.

• cluster.routing.allocation.allow_rebalance (Dynamic, string): Specifies when shard rebalancing is allowed.

  Valid values are:

  • always – Always allow rebalancing.
  • indices_primaries_active – Only allow rebalancing when all primaries in the cluster are allocated.
  • indices_all_active – Only allow rebalancing when all shards in the cluster are allocated.

  Default is indices_all_active.

• cluster.routing.allocation.cluster_concurrent_rebalance (Dynamic, integer): Allows you to control how many concurrent shard rebalances are allowed across a cluster. Default is 2.

• cluster.routing.allocation.balance.shard (Dynamic, floating point): Defines the weight factor for the total number of shards allocated per node. Default is 0.45.

• cluster.routing.allocation.balance.index (Dynamic, floating point): Defines the weight factor for the number of shards per index allocated on a node. Default is 0.55.

• cluster.routing.allocation.balance.threshold (Dynamic, floating point): The minimum optimization value of operations that should be performed. Default is 1.0.

• cluster.routing.allocation.balance.prefer_primary (Dynamic, Boolean): When set to true, OpenSearch attempts to evenly distribute the primary shards between the cluster nodes. Enabling this setting does not always guarantee an equal number of primary shards on each node, especially in the event of a failover. Changing this setting to false after it was set to true does not invoke redistribution of primary shards. Default is false.

• cluster.routing.allocation.rebalance.primary.enable (Dynamic, Boolean): When set to true, OpenSearch attempts to rebalance the primary shards between the cluster nodes. When enabled, the cluster tries to maintain the number of primary shards on each node, with the maximum buffer defined by the cluster.routing.allocation.rebalance.primary.buffer setting. Changing this setting to false after it was set to true does not invoke the redistribution of primary shards. Default is false.

• cluster.routing.allocation.rebalance.primary.buffer (Dynamic, floating point): Defines the maximum allowed buffer of primary shards between nodes when cluster.routing.allocation.rebalance.primary.enable is enabled. Default is 0.1.

• cluster.routing.allocation.disk.threshold_enabled (Dynamic, Boolean): When set to false, disables the disk allocation decider. This will also remove any existing index.blocks.read_only_allow_delete index blocks when disabled. Default is true.

• cluster.routing.allocation.disk.watermark.low (Dynamic, string): Controls the low watermark for disk usage. When set to a percentage, OpenSearch will not allocate shards to nodes with that percentage of disk usage. This can also be entered as a ratio value, like 0.85. Finally, this can also be set to a byte value, like 400mb. This setting does not affect the primary shards of newly created indexes but will prevent their replicas from being allocated. Default is 85%.

• cluster.routing.allocation.disk.watermark.high (Dynamic, string): Controls the high watermark. OpenSearch will attempt to relocate shards away from a node whose disk usage is above the defined percentage. This can also be entered as a ratio value, like 0.85. Finally, this can also be set to a byte value, like 400mb. This setting affects the allocation of all shards. Default is 90%.

• cluster.routing.allocation.disk.watermark.flood_stage (Dynamic, string): Controls the flood stage watermark. This is a last resort to prevent nodes from running out of disk space. OpenSearch enforces a read-only index block (index.blocks.read_only_allow_delete) on every index that has one or more shards allocated on the node and at least one disk exceeding the flood stage. The index block is released once the disk utilization falls below the high watermark. This can also be entered as a ratio value, like 0.85. Finally, this can also be set to a byte value, like 400mb. Default is 95%.

• cluster.info.update.interval (Dynamic, time unit): Sets how often OpenSearch should check disk usage for each node in the cluster. Default is 30s.

• cluster.routing.allocation.awareness.attributes (Dynamic, list): See Shard allocation awareness.

• cluster.routing.allocation.include.<attribute> (Dynamic, enum): Allocates shards to a node whose attribute contains at least one of the included comma-separated values.

• cluster.routing.allocation.require.<attribute> (Dynamic, enum): Only allocates shards to a node whose attribute contains all of the included comma-separated values.

• cluster.routing.allocation.exclude.<attribute> (Dynamic, enum): Does not allocate shards to a node whose attribute contains any of the included comma-separated values. The cluster allocation settings support the following built-in attributes.

  Valid values are:

  • _name – Match nodes by node name.
  • _host_ip – Match nodes by host IP address.
  • _publish_ip – Match nodes by publish IP address.
  • _ip – Match either _host_ip or _publish_ip.
  • _host – Match nodes by hostname.
  • _id – Match nodes by node ID.
  • _tier – Match nodes by data tier role.

• cluster.routing.allocation.awareness.force.<attribute>.values (Dynamic, list): See Forced awareness.

• cluster.routing.allocation.shard_movement_strategy (Dynamic, enum): Determines the order in which shards are relocated from outgoing to incoming nodes.

  This setting supports the following strategies:

  • PRIMARY_FIRST – Primary shards are relocated first, before replica shards. This prioritization may help prevent a cluster’s health status from going red if the relocating nodes fail during the process.
  • REPLICA_FIRST – Replica shards are relocated first, before primary shards. This prioritization may help prevent a cluster’s health status from going red when carrying out shard relocation in a mixed-version, segment-replication-enabled OpenSearch cluster. In this situation, primary shards relocated to OpenSearch nodes of a newer version could try to copy segment files to replica shards on an older version of OpenSearch, which would result in shard failure. Relocating replica shards first may help to avoid this in multi-version clusters.
  • NO_PREFERENCE – The default behavior in which the order of shard relocation has no importance.

• cluster.routing.search_replica.strict (Dynamic, Boolean): Controls how search requests are routed when search replica shards exist for an index, such as when index.number_of_search_replicas is greater than 0. This setting applies only when search replicas are configured for an index. When set to true, all search requests for such indexes are routed only to search replica shards. If search replicas are unassigned, the requests fail. When set to false, if search replicas are unassigned, requests fall back to any available shard. Default is true.

• cluster.allocator.gateway.batch_size (Dynamic, integer): Limits the number of shards sent to data nodes in a single batch to fetch any unassigned shard metadata. Default is 2000.

• cluster.allocator.existing_shards_allocator.batch_enabled (Static, Boolean): Enables batch allocation of unassigned shards that already exist on the disk, as opposed to allocating one shard at a time. This reduces memory and transport overhead by fetching any unassigned shard metadata in a batch call. Default is false.

• cluster.routing.allocation.total_shards_per_node (Dynamic, integer): The maximum combined total number of primary and replica shards that can be allocated to a single node. Default is -1 (unlimited). Helps distribute shards evenly across nodes by limiting the total number of shards per node. Use with caution because shards may remain unallocated if nodes reach their configured limits.

• cluster.routing.allocation.total_primary_shards_per_node (Dynamic, integer): The maximum number of primary shards that can be allocated to a single node. This setting is applicable only for remote-backed clusters. Default is -1(unlimited). Helps distribute primary shards evenly across nodes by limiting the number of primary shards per node. Use with caution because primary shards may remain unallocated if nodes reach their configured limits.

Cluster-level shard, block, and task settings

OpenSearch supports the following cluster-level shard, block, and task settings:

• action.search.shard_count.limit (Integer): Limits the maximum number of shards to be hit during search. Requests that exceed this limit will be rejected.

• cluster.blocks.read_only (Boolean): Sets the entire cluster to read-only. Default is false.

• cluster.blocks.read_only_allow_delete (Boolean): Similar to cluster.blocks.read_only, but allows you to delete indexes.

• cluster.no_cluster_manager_block (String): Configures the operations that are rejected when no cluster manager is active. Accepts one of the following three option:
  • all: Blocks all read and write requests to the cluster.
  • write: Blocks only write requests. Read requests can still be processed.
  • metadata_write: Metadata-related writes, such as updates to mappings or routing tables, are blocked, but normal document indexing can still be performed. Read and write requests are processed using the cluster state last received by the node. Because the node may be cut off from the rest of the cluster, this can lead to serving outdated information or returning only part of the data.

• cluster.max_shards_per_node (Integer): Limits the total number of primary and replica shards for the cluster. The limit is calculated as follows: cluster.max_shards_per_node multiplied by the number of non-frozen data nodes. Shards for closed indexes do not count toward this limit. Default is 1000.

• cluster.persistent_tasks.allocation.enable (String): Enables or disables allocation for persistent tasks.

  Valid values are:

  • all – Allows persistent tasks to be assigned to nodes.
  • none – No allocations are allowed for persistent tasks. This does not affect persistent tasks already running.

  Default is all.

• cluster.persistent_tasks.allocation.recheck_interval (Time unit): The cluster manager automatically checks whether persistent tasks need to be assigned when the cluster state changes in a significant way. There are other factors, such as memory usage, that will affect whether persistent tasks are assigned to nodes but do not otherwise cause the cluster state to change. This setting defines how often assignment checks are performed in response to these factors. Default is 30 seconds, with a minimum of 10 seconds being required.

Cluster-level search settings

OpenSearch supports the following cluster-level search settings:

• cluster.search.ignore_awareness_attributes (Boolean): Controls whether awareness attributes are considered during shard query routing. If true (default), the cluster ignores awareness attributes and uses Adaptive Replica Selection (ARS) to choose the optimal shard copy, reducing query response latency. Set this to false for routing decisions to prioritize awareness attributes instead of performance-based selection.

Cluster-level slow log settings

For more information, see Search request slow logs.

• cluster.search.request.slowlog.threshold.warn (Time unit): Sets the request-level slow log WARN threshold. Default is -1.

• cluster.search.request.slowlog.threshold.info (Time unit): Sets the request-level slow log INFO threshold. Default is -1.

• cluster.search.request.slowlog.threshold.debug (Time unit): Sets the request-level slow log DEBUG threshold. Default is -1.

• cluster.search.request.slowlog.threshold.trace (Time unit): Sets the request-level slow log TRACE threshold. Default is -1.

• cluster.search.request.slowlog.level (String): Sets the minimum slow log level to log: WARN, INFO, DEBUG, and TRACE. Default is TRACE.

Cluster-level index settings

For information about index-level index settings, see Cluster-level index settings.

Cluster-level coordination settings

OpenSearch supports the following cluster-level coordination settings. All settings in this list are dynamic:

• cluster.fault_detection.leader_check.timeout (Time unit): The amount of time a node waits for a response from the elected cluster manager during a leader check before deeming the check a failure. Valid values are from 1ms to 60s, inclusive. Default is 10s. Changing this setting to a value other than the default can result in an unstable cluster.

• cluster.fault_detection.follower_check.timeout (Time unit): The amount of time the elected cluster manager waits for a response during a follower check before deeming the check a failure. Valid values are from 1ms to 60s, inclusive. Default is 10s. Changing this setting to a value other than the default can result in an unstable cluster.

• cluster.fault_detection.follower_check.interval (Time unit): The amount of time that the elected cluster manager waits between sending follower checks to other nodes in the cluster. Valid values are 100ms and higher. Default is 1000ms. Changing this setting to a value other than the default can result in an unstable cluster.

• cluster.follower_lag.timeout (Time unit): The amount of time that the elected cluster manager waits to receive acknowledgements for cluster state updates from lagging nodes. Default is 90s. If a node does not successfully apply the cluster state update within this period of time, it is considered to have failed and is removed from the cluster.

• cluster.publish.timeout (Time unit): The amount of time that the cluster manager waits for each cluster state update to be completely published to all nodes, unless discovery.type is set to single-node. Default is 30s.

Cluster-level fault detection settings

OpenSearch supports the following cluster-level fault detection settings that control how nodes detect and respond to failures:

• cluster.fault_detection.follower_check.retry_count (Static, integer): Sets how many consecutive follower check failures must occur before the elected cluster manager considers a node to be faulty and removes it from the cluster. This setting controls fault detection sensitivity for follower nodes. Default is 3. Warning: Changing this setting from the default may cause your cluster to become unstable.

• cluster.fault_detection.leader_check.interval (Static, time unit): Sets how long each node waits between checks of the elected cluster manager. This controls the frequency of leader health checks performed by follower nodes. Default is 1s. Warning: Changing this setting from the default may cause your cluster to become unstable.

• cluster.fault_detection.leader_check.retry_count (Static, integer): Sets how many consecutive leader check failures must occur before a node considers the elected cluster manager to be faulty and attempts to find or elect a new cluster manager. This setting controls fault detection sensitivity for the cluster manager node. Default is 3. Warning: Changing this setting from the default may cause your cluster to become unstable.

• cluster.indices.tombstones.size (Static, integer): Sets the maximum number of index deletion tombstones to keep in the cluster state. Index tombstones prevent nodes that are not part of the cluster when a delete occurs from joining the cluster and reimporting the index as though the delete was never issued. To prevent the cluster state from growing too large, only the most recent deletions are kept. Default is 500. You can increase this value if you expect nodes to be absent from the cluster and miss more than 500 index deletions, though this is rare. Tombstones use minimal space, but very large values (like 50,000) are not recommended.

Cluster-level CAT response limit settings

OpenSearch supports the following cluster-level CAT API response limit settings, all of which are dynamic:

• cat.indices.response.limit.number_of_indices (Integer): Sets a limit on the number of indexes returned by the CAT Indices API. The default value is -1 (no limit). If the number of indexes in the response exceeds this limit, the API returns a 429 error. To avoid this, you can specify an index pattern filter in your query (for example, _cat/indices/<index-pattern>).

• cat.shards.response.limit.number_of_shards (Integer): Sets a limit on the number of shards returned by the CAT Shards API. The default value is -1 (no limit). If the number of shards in the response exceeds this limit, the API returns a 429 error. To avoid this, you can specify an index pattern filter in your query (for example, _cat/shards/<index-pattern>).

• cat.segments.response.limit.number_of_indices (Integer): Sets a limit on the number of indexes returned by the CAT Segments API. The default value is -1 (no limit). If the number of indexes in the response exceeds this limit, the API returns a 429 error. To avoid this, you can specify an index pattern filter in your query (for example, _cat/segments/<index-pattern>).

Cluster-level metadata settings

OpenSearch supports the following cluster-level metadata settings:

• cluster.metadata.<key> (Dynamic, varies): Adds cluster metadata in the "cluster.metadata.key": "value" format. This setting is useful for persisting arbitrary, infrequently changing information about the cluster, such as contact details or annotations, without creating a dedicated index. Important: User-defined cluster metadata is not intended to store sensitive or confidential information. The values are visible to anyone with access to the Get Cluster Settings API and are recorded in the logs.

Cluster-level remote cluster settings

OpenSearch supports the following remote cluster settings:

• cluster.remote.initial_connect_timeout (Dynamic, time unit): Sets the timeout period for establishing initial connections to remote clusters when the node starts. This prevents nodes from hanging indefinitely during startup if remote clusters are unavailable. Default is 30s.

• cluster.remote.<cluster_alias>.mode (Dynamic, string): Specifies the connection mode for a specific remote cluster. Valid values are sniff (discovers and connects to multiple nodes in the remote cluster) and proxy (connects through a single proxy address). Default is sniff.

• node.remote_cluster_client (Static, Boolean): Controls whether a node can act as a cross-cluster client and connect to remote clusters. Default is true. Set to false to prevent a node from connecting to remote clusters. Remote cluster requests must be sent to a node on which this setting is enabled.

• cluster.remote.<cluster_alias>.skip_unavailable (Dynamic, Boolean): Controls whether cross-cluster operations should continue when this specific remote cluster is unavailable. When set to true, the cluster becomes optional and operations will skip it if unreachable. When set to false, operations fail if this cluster is unavailable. Default is false.

• cluster.remote.<cluster_alias>.transport.compress (Dynamic, Boolean): Enables compression for transport communications with a specific remote cluster. When enabled, reduces network bandwidth usage but increases CPU overhead for compression operations. If unset, uses the global transport.compress setting as a fallback. Default is false.

• cluster.remote.<cluster_alias>.transport.ping_schedule (Dynamic, time unit): Sets the interval for sending application-level ping messages to maintain connections with a specific remote cluster. Setting to -1 disables pings for this cluster. If unset, uses the global transport.ping_schedule setting. Default is -1 (disabled).

• cluster.remote.<cluster_alias>.seeds (Dynamic, list): Applicable to sniff mode only. Specifies the list of seed nodes used to discover the remote cluster topology. These nodes are contacted initially to retrieve the cluster state and identify gateway nodes for ongoing connections.

• cluster.remote.<cluster_alias>.node_connections (Dynamic, integer): Applicable to sniff mode only. Sets the number of gateway nodes to which to maintain active connections in the remote cluster. More connections provide better availability but consume more resources. Default is 3.

• cluster.remote.node.attr (Static, string): Applicable to sniff mode only. Specifies a node attribute to filter nodes that are eligible as gateway nodes in remote clusters. When set, only remote cluster nodes with the specified attribute will be used for connections. For example, if remote cluster nodes have node.attr.gateway: true and this setting is set to gateway, only those nodes will be connected to for cross-cluster operations.

• cluster.remote.<cluster_alias>.proxy_address (Dynamic, string): Applicable to proxy mode only. Specifies the proxy server address for connecting to the remote cluster. All remote connections are routed through this single proxy endpoint.

• cluster.remote.<cluster_alias>.proxy_socket_connections (Dynamic, integer): Applicable to proxy mode only. Sets the number of socket connections to open to the proxy server for this remote cluster. More connections can improve throughput but consume more resources. Default is 18.

• cluster.remote.<cluster_alias>.server_name (Dynamic, string): Applicable to proxy mode only. Specifies the hostname sent in the TLS Server Name Indication (SNI) extension when TLS is enabled for the remote cluster connection. This must be a valid hostname according to TLS SNI specifications.