Enterprise Analytics Configuration REST API (2.1)

The maximum time to wait for tasks to be aborted, in milliseconds.

The memory budget for the active runtime. Accepts either an integer number of bytes or a byte string (e.g. 64MiB).

The maximum time (in seconds) to wait for a graceful stop of an active runtime.

The maximum time (in seconds) to wait for a graceful suspend of an active runtime.

Allow remote link redirects to loopback addresses.

The number of processed mutations after which the DCP state is broadcast to storage.

The maximum number of HTTP requests to queue pending ability to execute.

The number of threads to service HTTP requests.

How long (in seconds) to cache user authorizations before checking with the authorization provider.

How long to wait (in seconds) when receiving 401 from the parent process for credentials to succeed before halting. (0 == no wait, -1 == wait forever)

Whether the AWS credential provider should fetch credentials asynchronously in the background. If true, threads are less likely to block when credentials are loaded, but additional resources are used to maintain the provider.

AWS assuming role duration in seconds. Range from 900 seconds (15 minutes) to 43200 seconds (12 hours).

The amount of time (in seconds), relative to STS token expiration, at which cached credentials are considered close to stale and should be updated. Prefetch updates will occur between the specified time and the stale time of the provider.

The amount of time (in seconds), relative to STS token expiration, that the cached credentials are considered stale and must be updated.

Timeout (in seconds) for Azure client requests.

The specific bind address to use. If not set, the wildcard address is used.

If true, bind to the configured hostname instead of the wildcard address.

The interval (in seconds) at which to attempt to preemptively verify cloud access. If access is confirmed to be revoked, the process will halt. (0 == no preemptive refresh)

The length of time (in seconds) before the process will halt on failure to verify cloud access. (0 == wait forever)

The length of time (in seconds) allowed to assume we still have write access to cloud resources.

The waiting time (in milliseconds) if a requesting thread failed to acquire a token when the rate limit of cloud requests is exceeded. (min: 1, max: 5000)

The number of eviction events after which the eviction plan is re-evaluated.

The maximum number of read requests per second (default 1500, 0 means unlimited).

The maximum number of write requests per second (default 250, 0 means unlimited).

The interval (in minutes) at which to log cloud request statistics. Minimum is 1 minute. Note: enabling this logging may perturb the performance of cloud requests. (0 == disabled)

The waiting time (in seconds) to acquire an HTTP connection before failing the request. (default 120 seconds)

The time (in seconds) after which an idle cloud connection will be closed. (0 == unlimited idle)

Available from Enterprise Analytics 2.1.1.

The time (in seconds) after which a cloud connection will no longer be reused. (0 == unlimited lifetime)

Available from Enterprise Analytics 2.1.1.

The maximum number of HTTP connections to use concurrently for cloud requests per node. (default 1000)

The maximum number of HTTP connections allowed to wait for a connection per node. (default 10000)

The percentage of the total disk space that should be allocated for data storage when the selective caching policy is used. The remaining will act as a buffer for query workspace (i.e., for query operations that require spilling to disk). (default 80% of the total disk space)

The caching policy for cloud storage data. Supported values: - selective — Only cache data that is likely to be reused; sweeps disk when storage thresholds are exceeded. - eager — Cache all data eagerly. - lazy — Cache data on first access.

The disk monitoring interval time (in seconds) that determines how often the system checks for pressure on disk space when using the selective caching policy. (default 120 seconds)

The duration in minutes to consider an index is inactive. (default 360 or 6 hours)

The read timeout (in seconds) for the S3 synchronous client. -1 means the SDK default.

The S3 client type used for parallel downloads. Defaults to crt when no custom endpoint is configured, and java otherwise (computed at startup).

Enable round-robin DNS resolver for the S3 client.

The percentage of the used storage space at which the disk sweeper starts freeing space by punching holes in stored indexes or by evicting them entirely, when the selective caching policy is used. (default 90% of the allocated space for storage)

The write buffer size in bytes. (default 8MB, min 5MB)

The number of times to retry connecting to the cluster controller.

The maximum time (in seconds) to wait to collect DCP state from all nodes.

Enable/disable using array-indexes in queries.

Enable/disable batch point-lookups when running queries with secondary indexes.

Set the mode for cost based optimization.

Enable/disable the use of column min/max filters.

The COPY TO write buffer size in bytes. (default 8MB, min 5MB).

Rewrite disjunctions to joins in query plans.

Available from Enterprise Analytics 2.1.1. The default is false for clusters provisioned from 2.1.1 onwards. Clusters upgraded to 2.1.1 have this set to true to preserve pre-upgrade behavior.

The number of disjunctions after which a hash-based approach is used for evaluating OR operations. -1 disables the hash-based approach.

Available from Enterprise Analytics 2.1.1.

The maximum number of common expressions to extract.

Available from Enterprise Analytics 2.1.1. The default is 100 for clusters provisioned from 2.1.1 onwards. Clusters upgraded to 2.1.1 have this set to 2147483647 (unlimited) to preserve pre-upgrade behavior.

Enable pushdown of field accesses to the external data-scan operator.

The memory budget for an external scan operator instance in a partition. Accepts either an integer number of bytes or a byte string (e.g. 8KiB, 1MiB).

The page size for computation. Accepts either an integer number of bytes or a byte string (e.g. 32KiB).

The memory budget for a group by operator instance in a partition. Accepts either an integer number of bytes or a byte string (e.g. 32MiB).

The memory budget for a join operator instance in a partition. Accepts either an integer number of bytes or a byte string (e.g. 32MiB).

Maximum occurrences of a variable allowed in an expression for inlining.

The minimum memory budget for a group by operator instance in a partition. Accepts either an integer number of bytes or a byte string (e.g. 512KiB, 32MiB).

The minimum memory budget for a join operator instance in a partition. Accepts either an integer number of bytes or a byte string (e.g. 512KiB, 32MiB).

The minimum memory budget for a sort operator instance in a partition. Accepts either an integer number of bytes or a byte string (e.g. 512KiB, 32MiB).

The minimum memory budget for a window operator instance in a partition. Accepts either an integer number of bytes or a byte string (e.g. 512KiB, 32MiB).

Avoid costly O(N^2) expression optimization passes (e.g. constant folding, CSE) on large IN-list queries when the number of OR/AND function arguments is greater than or equal to the specified value.

Available from Enterprise Analytics 2.1.1. The default is 200 for clusters provisioned from 2.1.1 onwards. Clusters upgraded to 2.1.1 have this set to 2147483647 (unlimited) to preserve pre-upgrade behavior.

Enable/disable select order list.

The degree of parallelism for query execution. Zero means to use the storage parallelism as the query execution parallelism, while other integer values dictate the number of query execution parallel partitions. The system will fall back to use the number of all available CPU cores in the cluster as the degree of parallelism if the number set by a user is too large or too small.

Set the mode for forcing the shape of the query plan (e.g., zigzag, leftdeep, rightdeep).

A percentage of the job's required memory to be added to account for runtime memory overhead.

The memory budget for a sort operator instance in a partition. Accepts either an integer number of bytes or a byte string (e.g. 32MiB).

Enables/disables full parallel sort.

The memory budget for a window operator instance in a partition. Accepts either an integer number of bytes or a byte string (e.g. 32MiB).

The number of seconds to wait for COPY TO KV bucket WaitUntilReady before timing out.

The factor to multiply by the number of cores to determine maximum query concurrent execution level.

The integer percentage of DCP connection buffer size at which to acknowledge bytes consumed to the DCP producer as part of flow control.

The maximum attempt time to connect to a DCP channel in seconds.

The maximum time to wait to connect to a DCP channel in seconds.

DCP connection buffer size in bytes.

Default: Computed at startup — 10 MiB divided by the number of IO devices on the node when the JVM maximum heap is less than 8 GiB; otherwise 1% of the JVM maximum heap divided by the number of IO devices. A plain integer value in bytes may be specified to override this.

Directs KV to snappy compress all values before sending to Analytics. This should only be enabled in specific circumstances, as it will increase memory and CPU use on data nodes. Only applicable if snappy compression support is enabled (dcpSupportSnappyCompression).

The number of milliseconds before an empty DCP queue will trigger a flush of any pending incoming mutations to storage. (0 == flush instantly, negative value == never flush)

The maximum number of DCP connections to allow to each KV node for an ingestion (cluster-wide).

The minimum amount of work to allow per DCP connection for each ingestion connection. Work is (vbuckets * streams), where one stream is used for each unique KV collection being ingested. When allocating connections, it will be attempted to keep work per connection above this number.

The DCP NOOP interval (in seconds) used for dead connection detection.

Whether the DCP client should include the purge sequence number in stream requests, if supported by the producer.

Support receiving values from KV via DCP in snappy compressed form.

The maximum amount of work to target per DCP connection for each ingestion connection. Work is (vbuckets * streams), where one stream is used for each unique KV collection being ingested. The larger this number, the fewer DCP connections that will be attempted to be used.

The delay (in seconds) to wait for graceful shutdown due to deadlocked threads, before halting.

The frequency (in seconds) to scan for deadlocked threads.

GCS impersonating service account duration in seconds. Range from 60 seconds (1 minute) to 3600 seconds (1 hour).

HTTP status code returned by the health check endpoint (GET :8095/api/v1/health) when the node is being rebalanced out of the cluster. This should indicate failure to upstream load balancers so that traffic is no longer routed to the node.

HTTP status code returned by the health check endpoint (GET :8095/api/v1/health) when the node is healthy and able to serve traffic. This should indicate success to upstream load balancers.

The timeout (in seconds) to wait for remaining partitions to complete bootstrapping after at least one has bootstrapped, before aborting ingestion and triggering a reconnect.

The number of historical jobs to retain in memory.

The maximum number of jobs to queue before rejecting new jobs.

The maximum nesting depth for JSON objects. The depth is a count of objects and arrays that have not been closed, { and [ respectively.

The maximum length of a JSON document. Accepts either an integer number of bytes or a byte string (e.g. 10MiB). 0 means no limit.

The maximum length of a JSON field name. Accepts either an integer number of bytes or a byte string.

The maximum length of a JSON number. Accepts either an integer number of bytes or a byte string. 0 or less means no limit.

The maximum length of a JSON string value. Accepts either an integer number of bytes or a byte string. 0 or less means no limit.

The maximum number of JSON tokens in a JSON object (0 or less is no limit). A token is a single unit of input, such as a number, a string, an object start or end, or an array start or end.

JVM arguments to pass to the Analytics Driver. The default is undefined (null).

Note that JVM arguments are generally not secure, and are exposed by cbcollect_info and the System Event log. To pass arguments opaquely, you may use Java command-line argument files.

Maximum number of bytes fetched by a Kafka consumer in one request.

The maximum number of links that can be active at any given time.

The number of redirects to follow when communicating with the local cluster via HTTP(S).

The number of redirects to follow when communicating with a remote link cluster via http for remote links not configured to prevent redirects.

The maximum accepted web request size in bytes.

How long (in seconds) to wait for the metadata node to register with the cluster controller.

The number of network buffers per thread.

The number of network I/O threads.

Attempt to perform delta recovery when a node is recovered after a failover.

The delay (in seconds) before a node is ejected from the cluster during rebalance from the time that the health check API on that node will start reporting an error response. This is to allow adequate time for the node to removed from consideration by clients utilizing the health check API, to avoid failures on rebalance out; 0 == means no delay.

The maximum time (in seconds) to wait for a remote link connection to establish. A value of 0 disables timeout; a value of -1 sets timeout to the system default.

The frequency at which permissions should be reevaluated for remote buckets/collections that may have been lost/restored for remote cluster versions at 7.0 or later; a value of 0 disables preemptive reevaluation.

The maximum time (in seconds) to wait after establishing the connection for remote links; the maximum time of inactivity between two data packets. A value of 0 disables timeout; a value of -1 sets timeout to the system default.

The number of times to retry on a remote link credential / config validation connection failure, i.e. on CREATE or ALTER LINK (0 == don't retry).

The length of time after which the remote storage size metric is recomputed.

The number of seconds to wait before declaring a replica inactive while it is catching up.

The number of seconds to wait for a failed replica to recover before suspending new writes.

The number of seconds to wait before sending a replicas report.

The maximum number of archived requests to maintain.

The frequency (in seconds) at which to debug-log serialization replacements.

The maximum number of open files in the buffer cache.

The page size for pages in the buffer cache. Accepts either an integer number of bytes or a byte string (e.g. 128KiB).

The maximum time in milliseconds to wait for acquiring a buffer from the column buffer pool.

The maximum memory allocated to the column buffer pool. Accepts either an integer number of bytes or a byte string (e.g. 512MiB).

Default: 3% of the JVM maximum heap size, or 100 MiB for JVM heaps below ~3.3 GiB (computed at startup). A value of 0 restores the computed default.

The maximum number of buffers in the column buffer pool.

The size of each buffer in the column buffer pool. Accepts either an integer number of bytes or a byte string (e.g. 4KiB).

The default block compression scheme for storage (e.g., snappy, none).

The number of bytes written before each disk force (fsync). Accepts either an integer number of bytes or a byte string (e.g. 16MiB).

The default storage format (either row or column).

The I/O scheduler for LSM flush and merge operations (e.g., greedy, async).

The maximum number of columns in the zero segment (page-zero). Accepts either an integer or a byte string.

The maximum number of concurrently executed flushes per partition (0 means unlimited).

The maximum number of concurrently executed merges per partition (0 means unlimited).

The maximum number of scheduled merges per partition (0 means unlimited).

The memory usage threshold (as a fraction) at which memory components should be flushed.

The maximum number of scheduled flush operations. 0 means that the value will be calculated as the number of partitions.

The number of memory components to be used per LSM index.

The page size for pages allocated to memory components. Accepts either an integer number of bytes or a byte string (e.g. 128KiB).

The implementation to use for writing page-zero in storage components. Possible values: default, sparse, adaptive.

The maximum disk write rate for each storage partition. 0 or less disables the limit. Accepts either an integer number of bytes per second or a byte string (e.g. 100MiB).

The maximum time (in seconds) to wait for a response for streaming requests; the maximum time of inactivity between two data packets. A value of 0 disables timeout; a value of -1 sets timeout to the system default.

The frequency (in seconds) at which to log diagnostic thread dumps.

The log level at which to emit diagnostic thread dumps (e.g., DEBUG, INFO, WARN).

A comma-separated list of trace categories to enable for diagnostic tracing. The default is undefined (null — no categories enabled).

JVM arguments to pass to the Analytics Driver. The default is undefined (null).

Node-specific JVM arguments are appended to service-level JVM arguments. If the same JVM argument appears in both the service-level arguments and the node-specific arguments, the node-specific argument takes priority.

Note that JVM arguments are generally not secure, and are exposed by [cbcollect_info] and the [System Event] log. To pass arguments opaquely, you may use Java command-line argument files.

The size of memory allocated to the disk buffer cache. The value should be a multiple of the buffer cache page size. Accepts either an integer number of bytes or a byte string (e.g. 512MiB, 2GiB).

Default: 1/4 of the JVM maximum heap size (computed at startup).