How to configure graph schema.


schema.config().option(arg).[ set(value) | unset(value) | get() | exists() | describe() ]

Schema can be configured per graph using the config() command. An option and value can be set() or unset(). An option's value can be retrieved with the get() command. Whether or not the option is configured can be discovered with the exists() command. The describe() command returns a value if the option has been set manually.

Table 1. Graph-specific optionsGraph-specific options are preceded by graph. For example, graph.schema_mode.
Option argument Setting Example Description Default
allow_scan true Setting to allow costly graph scan queries. true
schema_mode Production Set mode to Production or Development. Development
default_property_key_cardinality multiple Set the cardinality that will be used by default unless otherwise specified. single
tx_autostart true Set whether transactions are started automatically or must be manually opened. false
Table 2. TraversalSource-specific optionsTraversalSource-specific options are preceded by graph.traversal_sources.* where * must be a specified traversal source such as the graph traversal g. For example, graph.traversal_sources.g.type. The most common TraversalSource is the graph traversal g.
Option argument Setting Example Description Default
evaluation_timeout PT10S (10 seconds) or "1500 ms" Maximum time to wait for a traversal to evaluate - this will override other system level settings for the current TraversalSource. 0 days
restrict_lambda false Prevent the use of lambdas with this TraversalSource. A particular traversal source can be identified. true
type read-only Specify type of TraversalSource. A particular traversal source can be identified. default
Important: Setting a timeout value of greater than 1095 days (maximum integer) can exceed the limit of a graph session. Starting a new session and setting the timeout to a lower value can recover access to a hung session. This caution is applicable for all timeouts: evaluation_timeout, system_evaluation_timeout, analytic_evaluation_timeout, and realtime_evaluation_timeout
Table 3. Transaction-specific optionsTransaction-specific options are preceded by graph.tx_groups.* where * must be specified as a transaction group or default. For example, graph.tx_groups.default.read_only will make all transactions which aren't explicitly named read_only, whereas graph.tx_groups.myTxGroup.read_only would apply only to transactions which are given the group name myTxGroup.
Option Setting Example Description Default
authenticated_user test_user The username to use as the current user for a transaction. ANONYMOUS_USER
cache true Cache retrievals and data store calls within a transaction in transaction-level caches. This setting provides a restricted type of isolation within a transaction (concurrent modifications in other transactions aren't visible and result sets remain consistent between calls) and can improve performance at the expense of additional memory consumption. true
deep_profiling true Enable CQL tracing for profile() in queries. Very costly profiling. false
internal_vertex_verify true Set whether a transaction should verify that vertices for internally provided vertex ids (autogenerated vertex ids) actually exist. false
external_vertex_verify false Set whether a transaction should verify that vertices for externally provided vertex ids (custom vertex ids) actually exist. true
logged_batch true Use a logged batch when committing changes. This guarantees that all mutations will eventually occur at the expense of performance. false
max_mutations 5000 The maximum number of vertices, properties and edges (cumulatively) that may be added or removed in a single transaction. 10000
max_profile_events 5 The maximum number of profiling events to report for an individual traversal step. Restricting the number of reported events makes output manageable, but can hide important information. 10
prefetch true Sets whether the query executor should asynchronously pre-fetch data based on its expected execution of the traversal prior to the data being requested. This can reduce transaction latency but can cause throughput to worse. true
read_only true Set whether a transaction is read-only. false
read_consistency ALL Specify the consistency level for read operations of a transaction. ONE
single_thread true Set whether a transaction is only accessed by a single thread. false
thread_bound true Set whether a transaction is bound to a particular thread. false
transaction_timestamp The timestamp at which all mutations of this transaction are persisted. Instant.EPOCH
verify_unique false Set whether transactions should ensure that uniqueness constraints are enforced. true
vertex_cache_size 4000 Maximum size of the transaction-level cache of recently-used vertices 20000l
vertex_dirty_size This is a performance hint for write-heavy, performance-sensitive transactional workloads. If set, it should roughly match the median vertices modified per transaction. 32
write_consistency ANY Specify the consistency level for write operations of a transaction QUORUM


Configure a graph. Options can be set, unset, or get (retrieve the value).


Set the current graph to disallow full graph scans in the currently aliased graph.
To retrieve all traversal sources that have been set, use the get() command with the traversal source type option:
resulting in a list of values for the option that have been manually set:
Set the default write consistency for a transaction to ALL in the currently aliased graph.
Get the current write consistency for a transaction in the currently aliased graph.
To confirm that an option setting has been set manually, use the exists() command:
This command will return:
if the setting has been set to a value, otherwise it returns false.
To enable CQL tracing during traversal query profiling, set the deep_profiling() option:
To verify that external vertex ids exist, set the external_vertex_verify() option:
If this setting is true, then a vertex will not be returned if it doesn't exist. However, if external_vertex_verify() is set to false, then a vertex will be returned even if the vertex does not exist given an id. Applications should ensure that vertices exist using the exists() method for expected behavior.
To retrieve a list of configuration options that have been set, use the describe() command:
resulting in a list of all options that have been manually set:
graph.tx_groups.default.write_consistency: ALL
graph.allow_scan: False

There are some configuration options for which the default (for example, values are not explicitly set) is determined by using the value of other configuration options. For instance, if allow_scan is not explicitly set, the default value is true if schema_mode is set to Development, but false if the schema_mode is set to Production. These configuration options are not linked to the default settings, leading to potentially misleading information when using schema.config().get() to discover the setting value because the default value is displayed rather than a set value.

To set restrict_lambda to FALSE in order to test lambda functions (only appropriate for non-production systems):
Full graph scan settings are as follows:
setting schema mode scans allowed
dse.yaml schema_mode:Production Production no
dse.yaml schema_mode:Development Development yes
graph.schema_mode:Production Production no
graph.schema_mode:Development Development no
graph.allow_scan:true Production yes
graph.allow_scan:true Development yes