7.7. Runtime Configuration¶
GeoMesa uses system properties for various runtime configuration options. As a convenience, properties
can be specified in an XML file instead of the command line. When run, GeoMesa will load
geomesa-site.xml from the classpath (if available), and use any properties configured there.
To configure a GeoMesa binary distribution, place
geomesa-site.xml in the
For GeoServer, place the file under
geoserver/WEB-INF/classes. For other environments,
ensure the file is available at the root level of the classpath.
Each tools distribution contains a template file with the default settings at
conf/geomesa-site.xml.template. Do not modify this file directly as it is never read;
instead copy the desired configurations into
By default, system properties set through command line parameters will take precedence over the
configuration file. If you wish a configuration item to always take precedence, even over command
line parameters, change the
<final> tag to true.
Configuration properties with empty values will not be applied, you can change this by marking a property as final.
7.7.1. Common Properties¶
These properties apply to all GeoMesa implementations. Additional properties for different back-end databases can be found in the chapters for each one.
Sets the IPC format version for Arrow-encoded responses. This is expected to be a valid Arrow version,
0.10. The Arrow IPC format changed slightly starting with version
This property specifies the fully-qualified class name of an audit provider implementation to use. For more information, see Query Auditing.
This property allows for adding GeoMesa converter configurations to the environment. It can be set to a comma-separated list of arbitrary URLs. For more information on converters, see GeoMesa Convert.
This property allows for adding files to the classpath. It should be set to a colon-separated list of file paths. This is useful for getting scripts onto the classpath for use by map-reduce ingest jobs.
This property controls the batch size used for running distributed density (heatmap) queries. It needs to be set on the client making the request (e.g. GeoServer). If a query is closed or cancelled before completion, the batch size will determine how long the distributed scan will keep running before seeing the cancellation.
The property controls the length of time a data store will wait to acquire a distributed lock before performing
schema operations (
removeSchema). As GeoMesa is often run in parallel,
acquiring a distributed lock among different processes prevents metadata corruption that may result from multiple
threads altering the schema simultaneously. The timeout is specified as a duration, e.g.
1 minute or
30 seconds, with a default value of
This property can be used to check for version mismatches in the distributed classpath. When enabled, GeoMesa will throw an exception if it detects a major version discrepancy between the local classpath and the distributed classpath (e.g. HBase region servers or Accumulo tablet servers), as this will generally cause queries to fail. If not enabled, then classpath errors will not be detected proactively, and will likely result in runtime exceptions.
This property controls the default implementation used for generating IDs for simple features,
PROVIDED_FID hint is not set in the feature. It should be set to
the fully-qualified class name for a class implementing
GeoMesa includes an implementing class
org.locationtech.geomesa.utils.uuid.Z3FeatureIdGenerator, which creates
a unique feature ID partially based on the default time and geometry for the feature. This class is used by
default if nothing else is specified, and should work well for most situations.
Evaluating a filter of the form
name = 'john' OR name = 'jane' OR name = 'doe' can be slow if the number
of clauses is high. GeoMesa will optimize such filters by turning them into a hash lookup instead of a sequential
comparison. This property controls the threshold for switching to a hash lookup. By default, the threshold is
Note that for datastores with distributed filtering (e.g. HBase and Accumulo), this property needs to be set on the distributed processing nodes.
This property controls how long query filters will be cached in memory in remote processes (i.e. HBase region servers
and Accumulo tablet servers). For repeated queries, caching the filter can improve query times. However, complex
filters can require a substantial amount of memory overhead. The expiry is specified as a duration, e.g.
10 minutes or
1 hour. The default is
Note that to take effect, the property must be set on each region or tablet server.
This property controls how GeoMesa calculates the size of a result set (e.g.
By default, GeoMesa will estimate the size of a result set using statistics. This will provide a
rough estimate very quickly. Some applications rely on knowing the exact size of a result set up
front, so estimates will cause problems. To force GeoMesa to calculate the exact size of a result
set, you may set this property to
true. You may also override this behavior on a per-query basis
by using the query hint
The GeoTools API around
getCount indicates that a
maxFeatures setting should be respected.
QueryHints.EXACT_COUNT is true and maxFeatures is lower than this setting,
GeoMesa will run the query to determine how many records are in the result set.
Otherwise, GeoMesa will use the Stats API and respect the counting setting. The default for this setting is 1000.
This property controls the maximum number of coordinates that will be allowed in a geometry. During deserialization, it is possible that corrupted data will cause the length of a geometry to be incorrect, which can lead to attempting to allocate space for a large number of coordinates. This property can be used to set an upper limit on the space that will be allocated. By default there is no max length.
This property controls the maximum level of geometry collections recursively containing other geometry collections. During deserialization, it is possible that corrupted data will cause the type of a geometry to be incorrect, which can lead to creating many nested geometry collections in a recursive loop, causing a stack overflow. This property can be used to set an upper limit on the level of recursion. By default 3 levels of recursion are allowed.
This property controls how query geometries will be handled with respect to the anti-meridian. Acceptable values are
spatial4j (the default) will use the Spatial4J library, which will interpret a
geometry with a segment spanning more than 180 degrees of longitude as being inverted around the anti-meridian. To
prevent a geometry from being inverted, add way-points every 180 degrees.
none will interpret geometries
literally. In this case, to query around the anti-meridian, use an OR filter or a geometry collection.
As an example, the following filters both specify a 2-degree area around the anti-meridian:
// spatial4j processing "intersects(geom, 'POLYGON((-179 90, 179 90, 179 -90, -179 -90, -179 90))')" // no processing "intersects(geom, 'MULTIPOLYGON(((-179 90, -180 90, -180 -90, -179 -90, -179 90)),((179 90, 180 90, 180 -90, 179 -90, 179 90)))')"
While the following filters both specify a 358-degree globe-spanning polygon:
// spatial4j processing "intersects(geom, 'POLYGON((-179 90, 0 90, 179 90, 179 -90, 0 -90, -179 -90, -179 90))')" // no processing "intersects(geom, 'POLYGON((-179 90, 179 90, 179 -90, -179 -90, -179 90))')"
Controls the max length of an
ilike predicate that will be parsed by GeoMesa for attribute index queries.
Case-insensitive matches must be enumerated for each possible case, which will result in exponentially
increasing query ranges. The default value is
10, which will result in 1024 ranges.
Controls the batch size for local ingests via the command-line tools. By default, feature writers will be flushed every 20,000 features.
This property controls how often simple feature type metadata is read from the underlying data store.
updateSchema on a data store will not show up in other instances until the metadata
cache has expired. The expiry is specified as a duration, e.g.
10 minutes or
1 hour. The default
This property controls how scans against multiple, partitioned tables are executed. By default scans will be
executed sequentially. If set to
true, they will be executed in parallel. See Configuring Partitioned Indices
for details on partitioning.
This property controls how GeoMesa performs query planning. By default, GeoMesa uses heuristics to determine the
best index for a given query. Alternatively, this property may be set to
stats to use cached data statistics
and cost-based query planning. This may also be overridden on a per-query basis using the query hint
set to either
org.locationtech.geomesa.accumulo.index.QueryPlanner.CostEvaluation.Index. See Query Planning
for more details on query planning strategies.
In addition to
geomesa.query.decomposition.bits sets a
lower threshold on the size of the envelopes. It must be between 1 and 63, inclusive. See the Wikipedia article
on GeoHashes for the approximate spatial extent
of a given number of bits.
GeoMesa creates scan ranges based on the spatial predicates in a query. For complex spatial predicates, GeoMesa will decompose the geometry into smaller, rectangular envelopes, which avoids scanning over rows which don’t intersect the geometry. This behavior can be controlled through two properties.
geomesa.query.decomposition.multiplier controls the maximum number of envelopes that a geometry will be
decomposed into. If set below 2, no decomposition will be performed and instead the geometry envelope will be used.
GeoMesa attempts to process input filters in order to determine the best query plan for a given predicate. However,
since queries can be arbitrarily complex, this processing can potentially take a significant amount of time.
geomesa.query.processing.or.threshold sets a threshold for the complexity of an OR filter that
will be considered, based on the permutations of the filter. For example, the filter
A OR B OR C has three
(A OR B) AND (C OR D) has four permutations.
By default complex OR predicates will not be considered, which is suitable for most queries.
This property can be used to prevent long-running queries from overloading the system. When set,
queries will be closed after the timeout, even if not all results have been returned yet. The
timeout is specified as a duration, e.g.
1 minute or
This property will prevent full-table scans from executing. A full-table scan is any query that can’t be
constrained down using a search index, and thus requires scanning the entire data set. With large data sets,
such a scan can last a long time and be resource intensive. The property is specified as a Boolean, i.e.
For more granularity, it is also possible to specify the full-table scan behavior for individual schemas
replaced with the schema name (e.g. “gdelt”). Properties set for an individual schema will take precedence
over the globally-defined behavior.
This property works in conjunction with
geomesa.scan.block-full-table, above. If a query puts a reasonable limit
on the number of features that are returned (through the use of
maxFeatures), then it will not be blocked.
The property is specified as an integer. By default, a limit of 1000 or less is allowed.
This property provides a rough upper-limit for the number of row ranges that will be scanned for a single query. It is specified as a number. In general, more ranges will result in fewer false-positive rows being scanned, which will speed up most queries. However, too many ranges can take a long time to generate, and overwhelm clients, causing slowdowns. The optimal value depends on the environment.
This property controls how long simple feature serializers will be cached in memory. Lowering this value may
reduce the memory footprint of your application, at the cost of increased processing time. The expiry is specified
as a duration, e.g.
10 minutes or
1 hour. The default is
This property can be used to constrain the memory used to sort result sets. GeoMesa sorts results in the client
process memory, since the supported back-end databases don’t offer native ordering. To avoid having large
result sets exceed the client memory capacity, a memory threshold can be set. Once the size of a result set
exceeds this threshold, additional results will be written to disk and sorted there. Note that the actual memory
used may exceed the threshold, as the memory footprint calculation is an estimation. The threshold is specified
as a number of bytes, e.g.
1GB. The default is to always sort in memory.
Note that distributed Arrow queries will never use disk to sort, due to the nature of Arrow result batches. For supported back-ends, sorting on disk for Arrow queries can be achieved by disabling remote Arrow processing.
This property allows for adding GeoMesa simple feature type configurations to the environment. It can be set to a comma-separated list of arbitrary URLs. For more information on defining types, see Defining Simple Feature Types.
This property controls the batch size used for running distributed stat queries. It needs to be set on the client making the request (e.g. GeoServer). If a query is closed or cancelled before completion, the batch size will determine how long the distributed scan will keep running before seeing the cancellation.
This property controls whether GeoMesa will generate statistics for a given feature type during ingestion. It
is specified as a Boolean,
false. This property will be used when a feature type is first created,
if stats are not explicitly configured in the feature type user data or through the
data store parameter. See Configuring Cached Statistics for details on configuring the feature type. Note that
stats are currently only implemented for the Accumulo and Redis data stores.
This property allows for overriding strategy selection during query planning. It should specify the
full class name for a class implementing
The class must have a no-arg constructor.
By default GeoMesa will use heuristic-based query planning, which should work well for most situations. See Query Planning for more details on query planning strategies.