23. Upgrade Guide

This section contains general information on version upgrades, as well as version-specific changes that may affect the end user.

23.1. Compatibility Across Versions

23.1.1. Semantic Versioning

Starting with 2.0.0, GeoMesa is adhering to semantic versioning. Essentially, releases are broken down into major, minor and patch versions. For a version number like 2.0.1, 2 is the major version, 2.0 is the minor version, and 2.0.1 is the patch version.

Major version updates contain breaking public API changes. Minor version updates contain new or updated functionality that is backwards-compatible. Patch versions contain only backwards-compatible bug fixes. This delineation allows users to gauge the potential impact of updating versions.

Warning

Versions prior to 2.0.0 do not follow semantic versioning, and each release should be considered a major version change.

23.1.2. Compatibility

Semantic versioning makes API guarantees, but GeoMesa has several compatibility vectors to consider:

23.1.2.1. Data Compatibility

Data compatibility refers to the ability to read and write data written with older versions of GeoMesa. GeoMesa fully supports data written with version 1.2.2 or later, and mostly supports data written with 1.1.0 or later.

Note that although later versions can read earlier data, the reverse is not necessarily true. Data written with a newer client may not be readable by an older client.

Data written with 1.2.1 or earlier can be migrated to a newer data format. See Upgrading Existing Indices for details (note that this functionality is currently only implemented for Accumulo).

23.1.2.2. API Compatibility

The GeoMesa public API is not currently well defined, so API compatibility is only guaranteed at the GeoTools DataStore level. In the future, GeoMesa will clearly indicate which classes and methods are part of the public API. Non-public classes may change without warning between minor versions.

23.1.2.3. Binary Compatibility

Binary compatibility refers to the ability to have different GeoMesa versions in a single environment. An environment may be a single process or span multiple servers (for example an ingest pipeline, a query client, and an analytics platform). For data stores with a distributed component (HBase and Accumulo), the environment includes both the client and the distributed code.

GeoMesa requires that all JARs in an environment are the same minor version, and that all JARs within a single JVM are the same patch version.

23.1.2.4. Dependency Compatibility

Dependency compatibility refers to the ability to update GeoMesa without updating other components (e.g. Accumulo, HBase, Hadoop, Spark, GeoServer, etc). Generally, GeoMesa supports a range of dependency versions (e.g. Accumulo 1.6 to 1.9). Spark versions are more tightly coupled, due to the use of private Spark APIs.

23.1.2.5. Pre-Release Code

GeoMesa sometimes provides modules in an alpha or beta state. Although they share the overall GeoMesa version number, such modules should be considered pre-1.0, and are not guaranteed to provide any forwards or backwards compatibility across versions. Pre-release modules will be clearly marked in the documentation.

23.1.3. Compatibility Matrix

Major

Minor

Patch

Data

Y

Y

Y

API

N

Y

Y

Binary

N

N

Y

Dependencies

N

N

Y

23.2. Version 5.0.0 Upgrade Guide

23.2.1. Version Compatibility

GeoMesa 5.0.x is generally compatible with versions 4.0.x and 3.5.x across different environments. This means that it is possible to upgrade in parts; i.e. upgrade GeoServer to 5.0.0 but keep NiFi at 3.5.2. Please note that previously deprecated functionality (see below) may no longer work once any part of the environment is upgraded to 5.0.0.

23.2.2. Java Version

GeoMesa no longer supports Java 8. The Java ecosystem is slowly moving on from Java 8, and it is no longer possible to support Java 8 while staying up-to-date with dependencies and security patches. GeoMesa now supports Java versions 11 and 17.

23.2.3. GeoTools Upgrade

GeoTools has been updated to version 30.2. This version contains extensive package changes. In addition to the Ant migration script provided by GeoTools, GeoMesa provides a sed script to help migrate Scala projects. Use the following bash script to upgrade a project from an earlier version of GeoTools to version 30:

# run the following from the root directory of your project:
wget 'https://raw.githubusercontent.com/locationtech/geomesa/geomesa-5.0.0/build/remove-opengis.xml'
ant -f remove-opengis.xml
# for Scala projects, also run:
wget 'https://raw.githubusercontent.com/locationtech/geomesa/geomesa-5.0.0/build/gt-30-api-changes.sed'
find . -name "*.scala" -not -exec grep -q "import org.geotools.api.data._" {} \; -exec sed -E -i -f gt-30-api-changes.sed {} \;

23.2.4. HBase Versions

Support for HBase 1.4 has been dropped, as HBase 1.4 does not support Java 11.

23.2.5. Dependency Version Upgrades

The following dependencies have been upgraded:

  • accumulo 2.0.1 -> 2.1.2

  • aircompressor 0.21 -> 0.25

  • antlr 4.7.1 -> 4.7.2

  • arrow 11.0.0 -> 15.0.2

  • avro 1.11.1 -> 1.11.3

  • aws-java-sdk 1.11.179 -> 1.12.625

  • caffeine 2.9.3 -> 3.1.8

  • cassandra-driver 3.11.3 -> 3.11.5

  • com.clearspring.analytics 2.9.2 -> 2.9.8

  • com.fasterxml.jackson 2.14.1 -> 2.16.1

  • com.jayway.jsonpath 2.7.0 -> 2.9.0

  • com.typesafe:config 1.4.2 -> 1.4.3

  • commons-cli 1.2 -> 1.6.0

  • commons-codec 1.15 -> 1.16.0

  • commons-compress 1.22 -> 1.26.0

  • commons-configuration2 2.5 -> 2.10.1

  • commons-csv 1.9.0 -> 1.10.0

  • commons-dbcp2 2.6.0 -> 2.11.0

  • commons-io 2.8.0 -> 2.15.1

  • commons-lang3 3.8.1 -> 3.14.0

  • commons-pool2 2.6.1 -> 2.12.0

  • commons-text 1.10.0 -> 1.11.0

  • confluent 6.2.7 -> 6.2.9

  • cqengine 3.0.0 -> 3.6.0

  • org.apache.curator 4.3.0 -> 5.6.0

  • geotools 28.2 -> 30.2

  • gson 2.10 -> 2.10.1

  • guava 30.1-jre -> 32.0.0-jre

  • hadoop 2.10.2 -> 3.4.0

  • hbase 2.5.2 -> 2.5.8-hadoop3

  • httpclient 4.5.13 -> 4.5.14

  • httpcore 4.4.15 -> 4.4.16

  • io.netty 4.1.85.Final -> 4.1.106.Final

  • javax.measure:unit-api 2.0 -> 2.1.2

  • jcommander 1.78 -> 1.82

  • jedis 4.3.1 -> 5.1.0

  • kafka 2.8.2 -> 3.7.0

  • kryo 4.0.2 -> 4.0.3

  • orc 1.8.2 -> 1.9.3

  • org.eclipse.emf.common 2.15.0 -> 2.29.0

  • org.eclipse.emf.ecore 2.15.0 -> 2.35.0

  • org.eclipse.emf.ecore.xmi 2.15.0 -> 2.36.0

  • org.ehcache:sizeof 0.4.0 -> 0.4.3

  • parquet 1.12.3 -> 1.13.1

  • parboiled 1.3.1 -> 1.4.1

  • postgresql 42.5.1 -> 42.7.2

  • pureconfig 0.17.2 -> 0.17.4

  • saxon 11.4 -> 12.4

  • scala 2.12.17 -> 2.12.18

  • scala-parser-combinators 2.1.1 -> 2.3.0

  • scala-xml 2.1.0 -> 2.2.0

  • sedona 1.3.1-incubating -> 1.5.0

  • si.uom 2.0.1 -> 2.1

  • spark 3.3.1 -> 3.5.0

  • spring-security 5.8.0 -> 5.8.11

  • systems.uom 2.0.2 -> 2.1

  • tech.units:indriya 2.0.2 -> 2.2

  • tech.uom.lib 2.0 -> 2.1

  • xmlresolver 4.4.3 -> 5.2.2

  • zookeeper 3.5.10 -> 3.9.2

23.2.6. Deprecated Classes and Methods

The following classes have been deprecated and will be removed in a future version:

  • org.locationtech.geomesa.kafka.confluent.SchemaParser.GeoMesaAvroDeserializableEnumProperty

23.2.7. Removed Modules

The geomesa-metrics-ganglia module has been removed, and Ganglia is no longer supported as a destination for metrics.

As part of dropping support for HBase 1.x, the geomesa-hbase-distributed-runtime-hbase1, geomesa-hbase-server-hbase1 and geomesa-hbase-spark-runtime-hbase1 modules have been removed.

23.2.8. GeoMesa Convert OSM

The geomesa-convert-osm module has been relocated to https://github.com/geomesa/geomesa-convert-osm.

23.2.9. Partitioned PostGIS Prepared Statements

If not specified, prepared statements now default to true in the partitioned PostGIS data store. Prepared statements are generally faster on insert, and some attribute types (such as list-type attributes) are only supported through prepared statements.

23.3. Version 4.0.0 Upgrade Guide

23.3.1. Version Compatibility

GeoMesa 4.0.0 is focused on upgrading dependency versions and removing deprecated features, and only contains a few new features. To make upgrading easier, version 4.0.x is generally compatible with version 3.5.x across different environments. This means that it is possible to upgrade in parts; i.e. upgrade GeoServer to 4.0.0 but keep NiFi at 3.5.1. Please note that previously deprecated functionality (see below) may no longer work once any part of the environment is upgraded to 4.0.0.

23.3.2. Scala Versions

Scala 2.11 support has been removed, and Scala 2.13 support has been added (in addition to the existing Scala 2.12 support).

23.3.3. GeoTools/GeoServer Versions

GeoTools has been upgraded from 23.3 to 28.2. GeoServer has been upgrade from 2.17.3 to 2.22.2. JTS has been upgraded from 1.17.0 to 1.19.0.

As part of this upgrade, various GeoTools methods have changed in incompatible ways. Several classes, in particular Query and SimpleFeatureBuilder, have changed from accepting arrays to using varargs (variable arguments). Additionally, the various DataStore methods, such as DataStoreFinder, now require Map<String, ?> instead of Map<String, ? extends Serializable>.

23.3.4. Dependency Version Upgrades

The following high-level dependencies have been upgraded:

  • Apache Hadoop 2.8.5 -> 2.10.2

  • Apache Spark 2.4.7 -> 3.3.1

  • Apache Accumulo 2.0.0 -> 2.0.1

  • Apache HBase 1.4.12 -> 1.4.14, 2.2.3 -> 2.5.2

  • Apache Kafka 2.1.1 -> 2.8.2

  • Apache Arrow 0.16.0 -> 11.0.0

  • Apache Avro 1.8.2 -> 1.11.1

  • Apache Parquet 1.9.0 -> 1.12.3

  • Apache Orc 1.5.4 -> 1.8.2

  • Jedis 3.0.1 -> 4.3.1

  • Confluent 5.1.0 -> 6.2.7

  • Kryo 3.0.3 -> 4.0.2

  • Typesafe Config 1.3.3 -> 1.4.2

  • EJML 0.34 -> 0.41

  • Saxon 9.7.0-20 -> 11.4

For a full changelist of all dependencies, see the diff here.

23.3.5. Minimum Library Versions

Support for older versions of some libraries has been dropped. The following minimum versions are now required:

  • Apache Accumulo 2.0.0 (dropped support for 1.7, 1.8, 1.9, and 1.10)

  • Apache Spark 3.0 (dropped support for 2.4)

  • Apache Kafka 2.0 (dropped support for 0.10, 0.11, 1.0, and 1.1)

23.3.6. Removal of Deprecated Modules

The following deprecated modules were removed:

  • geomesa-bigtable

  • geomesa-kudu

  • geomesa-stream

  • geomesa-geojson

  • geomesa-web

  • geomesa-feature-nio

  • geomesa-convert-metrics-cloudwatch

  • geomesa-convert-metrics-ganglia

  • geomesa-convert-metrics-graphite

In addition, various other deprecated classes and methods were removed. To identify any code that requires changes, build your project against GeoMesa 3.5.1 and note any deprecation warnings generated by the compiler.

23.3.7. Package Changes

The following packages were moved, renamed or split in order to support Java 11 modules:

  • org.locationtech.geomesa.jobs.accumulo -> org.locationtech.geomesa.accumulo.jobs

  • org.locationtech.geomesa.spark.accumulo -> org.locationtech.geomesa.accumulo.spark

  • org.locationtech.geomesa.spark.hbase -> org.locationtech.geomesa.hbase.spark

  • org.locationtech.geomesa.arrow.vector (partial) -> package org.locationtech.geomesa.arrow.jts

  • org.locationtech.geomesa.parquet -> org.locationtech.geomesa.fs.storage.parquet

  • org.locationtech.geomesa.process (partial) -> org.locationtech.geomesa.process.wps

23.3.8. GeoMesa NiFi Changes

GeoMesa NiFi is now built against NiFi 1.19.1. The GeoMesa NARs and JARs have been renamed to include the Scala version (i.e. geomesa-datastore-services-nar_2.12-4.0.0.nar). The datastore-specific processors (e.g. PutGeoMesaHBase) have been removed in favor of the generic processors (e.g. PutGeoMesa). The recommended upgrade path is to first upgrade to GeoMesa NiFi 3.5.1, and replace all the datastore-specific processors in the flow. This will ensure that the flow is still valid after upgrading to GeoMesa NiFi 4.0.0. The geomesa-accumulo2-nar has been replaced with geomesa-accumulo20-nar, and there is an additional geomesa-accumulo21-nar for Accumulo 2.1 support.

23.3.9. Scan Range Changes

GeoMesa will now generate a more accurate number of ranges based on geomesa.scan.ranges.target. Users who have configured this property should verify their setting is still appropriate, especially if set to a large value. Setting geomesa.scan.ranges.recurse to 7 will restore the old behavior if needed.

23.3.10. Partitioned PostGIS Query Changes

GeoMesa will now ignore queries that encompass the entire world in the partitioned PostGIS data store. For more information, refer to Configuring Filter Optimizations.

23.4. Version 3.5.0 Upgrade Guide

23.4.1. Removal of Log4j

GeoMesa has been updated to ban all usages of log4j, to mitigate various CVEs present in that framework. In most cases, GeoMesa uses slf4j, and delegates to the logging framework of the runtime environment. However, this change impacts the JARs bundled with the command-line tools, which now ship with reload4j instead. Other environments using GeoMesa (i.e. GeoServer) must be hardened independently.

23.4.2. Kafka Serialization

The GeoMesa Kafka data store now supports a new serialization format, avro-native. This format uses Avro array and map types for List and Map type attributes, which makes it easier to read with standard Avro tools. Note that GeoMesa versions before 3.5.0 will not be able to consume topics written in this format.

23.4.3. Deprecated Modules

The following modules have been deprecated, and will be removed in a future version:

  • GeoMesa Bigtable

23.4.4. Dependency Updates

  • org.slf4j:slf4j-api: 1.7.25 -> 1.7.36

  • com.google.code.gson:gson: 2.8.1 -> 2.8.9

23.5. Version 3.3.0 Upgrade Guide

23.5.1. Scala Versions

GeoMesa NiFi NARs now ship with Scala 2.12 by default. This should be largely transparent to end-users, however any custom GeoMesa converter JARs used in NiFi and written in Scala will need to be compiled with Scala 2.12.

23.6. Version 3.2.0 Upgrade Guide

23.6.1. Scala Versions

GeoMesa now supports Scala 2.12. Scala 2.11 support has been deprecated and will be removed in a future version.

23.6.2. Spark Versions

GeoMesa now supports Spark 3.0 and 3.1. Support for Spark 2.3 and 2.4 has been deprecated and will be removed in a future version.

23.6.3. Dependency Updates

  • com.fasterxml.jackson: 2.9.10 -> 2.12.1

23.6.4. FileSystem Data Store Metadata Format Change

The metadata format for the FileSystem data store has been changed to support storing arbitrary key-value pairs. Any data written with version 3.2.0 or later will not be readable by earlier GeoMesa versions.

23.6.5. Lambda Data Store Binary Distribution Change

The Lambda data store binary distribution no longer contains the geomesa-accumulo-distributed-runtime JAR. This JAR is available in the Accumulo data store binary distribution.

23.6.6. StrategyDecider API Update

The org.locationtech.geomesa.index.planning.StrategyDecider API has been extended with an optional GeoMesaStats argument that enables stat-based strategy decisions. The old API method has been deprecated and will be removed in a future version.

23.6.7. Deprecated Modules

The following modules have been deprecated, and will be removed in a future version:

  • GeoMesa Kudu

  • GeoMesa Streaming (Camel integration)

  • GeoMesa Web

  • GeoMesa GeoJSON

23.6.8. Deprecated Arrow Output Options

The Arrow output options for providing cached dictionaries, returning multiple logical files, and running queries in two passes have been deprecated and will be removed in the next major version.

23.7. Version 3.1.0 Upgrade Guide

23.7.1. Maven Type of GeoServer Plugin Modules

All of the geomesa-*-gs-plugin artifacts have been changed to <type>pom</type>, since they did not contain any code. Any pom.xml references to them should be updated to use the correct type.

23.7.2. Avro Version Update

The version of Avro used by GeoMesa has been updated from 1.7.5 to 1.8.2. Avro serialized files should be compatible between versions, but compile and runtime dependencies may need to be updated if a project uses Avro and references GeoMesa.

23.7.3. Query Interceptors API Change

The query interceptors API has been expanded to support query guards. Any existing query interceptor implementations will continue to work, but may need to be re-compiled against the GeoMesa 3.1.0.

23.7.4. Dependency Updates

  • GeoTools: 23.0 -> 23.3

  • Avro: 1.7.5 -> 1.8.2

23.8. Version 3.0.0 Upgrade Guide

23.8.1. Removal of Deprecated Modules and Classes

GeoMesa 3.0.0 removes several lesser-used modules, as well as various obsolete classes and methods.

The modules removed are: geomesa-accumulo/geomesa-accumulo-compute, geomesa-accumulo/geomesa-accumulo-native-api, geomesa-accumulo/geomesa-accumulo-raster-distributed-runtime, geomesa-accumulo/geomesa-accumulo-raster, geomesa-accumulo/geomesa-accumulo-security, geomesa-accumulo/geomesa-accumulo-stats-gs-plugin, geomesa-convert/geomesa-convert-scripting, geomesa-convert/geomesa-convert-simplefeature, geomesa-hbase/geomesa-hbase-native-api, geomesa-metrics, geomesa-native-api, geomesa-spark/geomesa-spark-geotools, geomesa-blobstore/*, and geomesa-web/geomesa-web-data.

The classes and methods removed are detailed in GEOMESA-2284.

23.8.2. HBase 2 Support

GeoMesa 3.0.0 supports both HBase 1.4 and HBase 2.2. HBase 1.3 is no longer supported. HBase 2.0 and 2.1 are not officially supported, but may work in some cases.

There are now two separate modules for HBase filters and coprocessors - geomesa-hbase-distributed-runtime-hbase1 and geomesa-hbase-distributed-runtime-hbase2. The previous geomesa-hbase-distributed-runtime module has been removed. Users should install the distributed runtime corresponding to their HBase installation.

Similarly, there are now two separate modules for HBase Spark support - geomesa-hbase-spark-runtime-hbase1 and geomesa-hbase-spark-runtime-hbase2. The previous geomesa-hbase-spark-runtime module has been removed. Users should use the Spark runtime corresponding to their HBase installation.

23.8.3. Accumulo 2 Support

GeoMesa 3.0.0 supports both Accumulo 1.9 with Hadoop 2.8 and Accumulo 2.0 with Hadoop 3. Earlier versions of Accumulo are no longer supported, but may work in some cases.

There are now two separate modules for Accumulo Spark support - geomesa-accumulo-spark-runtime-accumulo1 and geomesa-accumulo-spark-runtime-accumulo2. The previous geomesa-accumulo-spark-runtime module has been removed. Users should use the Spark runtime corresponding to their Accumulo installation.

23.8.4. NiFi Processors

The GeoMesa NiFi processors have been updated to NiFi 11 and split out into separate nar files for each supported back-end database. Additionally, there are separate nar files for HBase 1.4/2.2 and Accumulo 1.9/2.0, respectively. The processor classes and configurations have also changed. See GeoMesa NiFi Bundle for details.

23.8.5. Dependency Updates

  • Apache Arrow: 0.10 -> 0.16

23.8.6. Apache Arrow Updates

As part of the upgrade to Apache Arrow 0.16, the geomesa-arrow modules have been refactored to simplify memory management and allocation. Some classes have been removed, and some interfaces have changed. This may impact anyone using the geomesa-arrow modules directly.

The Arrow IPC format changed in Arrow 0.15. Older clients may not be able to read Arrow-encoded results by default. To enabled the ‘legacy’ Arrow IPC format, set the system property geomesa.arrow.format.version to 0.10, or use the query hint ARROW_FORMAT_VERSION. See Arrow Encoding for details.

23.8.7. Converter Date Functions

The converter functions isoDate and isoDateTime have been updated to match the equivalent Java DateTimeFormatter pattern. isoDate has changed from yyyyMMdd to yyyy-MM-dd, while isoDateTime has changed from yyyyMMdd'T'HHmmss.SSSZ to yyyy-MM-dd'T'HH:mm:ss. The old patterns can still be referenced through basicDate and basicDateTime.

23.8.8. AuthorizationsProvider and AuditProvider API Change

The signature for org.locationtech.geomesa.security.AuthorizationsProvider#configure and org.locationtech.geomesa.utils.audit.AuditProvider#configure have changed slightly from void configure(Map<String, Serializable> params) to public void configure(Map<String, ? extends Serializable> params). Any classes implementing either of these interfaces will need to update their method signature. Any classes invoking these methods should not need to updated, as the new signature is compatible with the old one.

23.8.9. Accumulo Default Visibilities Removed

The Accumulo data store parameter geomesa.security.visibilities have been removed. Visibilities should be set per-feature, as described in Data Security.

23.9. Version 2.4.0 Upgrade Guide

23.9.1. GeoTools 21 and GeoServer 2.15

GeoMesa 2.4.0 is compiled against GeoTools 21.1 and GeoServer 2.15. This version of GeoTools contains package and class location changes to support Java 11. Due to the changes, GeoMesa will no longer work with older versions of GeoTools and GeoServer.

23.9.2. Configuration of Cached Statistics

GeoMesa 2.4.0 moves the configuration of cached stats from a data store parameter (where it has to be set every time) to the feature type user data (where it is set once at schema creation, and only changed through explicit schema updates). See Configuring Cached Statistics for more details.

Feature types that were created in prior versions will continue to behave as before, with the configuration determined by the data store parameter each time. The configuration can be set permanently through the updateSchema data store method or the update-schema CLI command.

23.9.3. Indexing of Timestamp Attributes

GeoMesa 2.4.0 fully supports indexing of java.sql.Timestamp attributes. In previous versions, timestamp attribute indices were not officially supported, however they did work in some cases. Any data that was written to a timestamp attribute index with an older version will no longer be readable by GeoMesa 2.4.0. To migrate old data, truncate the index table first, then re-write all existing records:

import org.geotools.data.{DataStoreFinder, Query, Transaction}
import org.locationtech.geomesa.index.geotools.GeoMesaDataStore
import org.locationtech.geomesa.utils.geotools.FeatureUtils

val params: java.util.Map[String, String] = ??? // data store connection parameters
val ds: GeoMesaDataStore[_] = DataStoreFinder.getDataStore(params).asInstanceOf[GeoMesaDataStore[_]]
val typeName: String = ??? // simple feature type name to update
val timestamps: Seq[String] = ??? // names of any timestamp-type attributes
val indices = ds.manager.indices(ds.getSchema(typeName)).filter(_.attributes.headOption.exists(timestamps.contains))
val writer = ds.getIndexWriterAppend(typeName, indices)
val features = ds.getFeatureReader(new Query(typeName), Transaction.AUTO_COMMIT)
try {
  while (features.hasNext) {
    FeatureUtils.write(writer, features.next(), useProvidedFid = true)
  }
} finally {
  features.close()
  writer.close()
}

23.9.4. NiFi Processor Changes

The GeoMesa NiFi processors have been refactored to support NiFi nar inheritance and as a first step towards supporting Java 11. Any existing processors will continue to work under the older version, as long as you don’t delete the old GeoMesa nar file. However, you will need to create new processors in order to upgrade to 2.4.0.

23.9.5. Distribution of Installation Bundles

As of GeoMesa 2.4.0, installation bundles (binary distribution and GeoServer plugin tar files) will no longer be hosted on Maven Central. They will continue to be available on GitHub and the Locationtech Maven Repository. Note that this only applies to large installation bundles; GeoMesa will continue to publish JAR files to Maven Central.

23.9.6. HBase GeoServer Plugin Installation

The GeoMesa HBase GeoServer plugin installation tar file has been updated to remove the shaded HBase client JARs. The appropriate client JARS for your HBase version now must be installed separately. See Installing GeoMesa HBase in GeoServer for details.

If desired, the shaded GeoMesa JAR is still available from Maven, as org.locationtech.geomesa:geomesa-hbase-gs-plugin_2.11 with the classifier shaded. However, this will likely be removed in the next major version release.

23.10. Version 2.3.0 Upgrade Guide

23.10.1. Default Query Planning Type

GeoMesa 2.3.0 changes the default query planning type from stat-based to heuristic-based. This will only affect the Accumulo data store, as other stores have not implemented statistics. To enable stat-based query planning, refer to Query Planning Type.

23.10.2. Immutable Simple Feature Types

GeoMesa 2.3.0 returns immutable objects from calls to getSchema. This allows for the re-use of SimpleFeatureType instances, which reduces overhead. In most cases, this will have no effect on end users, however note that mutable and immutable feature types will never be equals when compared directly.

In order to update a schema, or if mutability is desired for some other reason, call org.locationtech.geomesa.utils.geotools.SimpleFeatureTypes.mutable() to create a mutable copy. Java users can call org.locationtech.geomesa.utils.interop.SimpleFeatureTypes.mutable() instead.

23.10.3. FileSystem Storage API Changes

The FileSystem Storage API is still considered beta-level software, and has been updated in this release. The DataStore API has not changed, however the internal class interfaces have changed in this release, potentially requiring changes in user code.

In addition, the format used to store metadata files has been updated, so older versions of GeoMesa will not be able to read metadata created with this version.

23.10.4. Deprecated Modules

The following modules have been deprecated, and will be removed in a future version:

  • GeoMesa Raster

  • GeoMesa Native API

  • GeoMesa Blob Store

  • GeoMesa Metrics

23.11. Version 2.2.0 Upgrade Guide

23.11.1. GeoTools 20 and GeoServer 2.14

GeoMesa 2.2.0 is compiled against GeoTools 20.0 and GeoServer 2.14. This version of GeoTools upgrades JTS from 1.14 to 1.16, which includes a transition of the project to Locationtech. The new version of JTS renames the packages from com.vividsolutions to org.locationtech.jts. Due to the package renaming, GeoMesa will no longer work with older versions of GeoTools and GeoServer.

Warning

GeoMesa 2.2.0 requires GeoTools 20.x and GeoServer 2.14.x.

23.11.2. Accumulo DataStore GeoServer Installation

When using GeoServer, the GeoMesa Accumulo data store now requires Accumulo client JARs 1.9.2 or later. This is due to classpath conflicts between earlier Accumulo clients and GeoServer 2.14. Fortunately, newer Accumulo clients can talk to older Accumulo instances, so it is only necessary to upgrade the client JARs in GeoServer, but not the entire Accumulo cluster.

23.12. Version 2.1.0 Upgrade Guide

23.12.1. Converter Updates

The GeoMesa converter API has been updated and simplified. The old API has been deprecated, and while custom converters written against it should still work, users are encouraged to migrate to org.locationtech.geomesa.convert2.SimpleFeatureConverter. A compatibility bridge is provided so that all converters registered with either the new or old API will be available to both.

Converter definitions should continue to work the same, but some invalid definitions may start to fail due to stricter configuration parsing.

23.12.1.1. XML Converter Namespaces

XML parsing is now namespace-aware. This shouldn’t affect most operations, but any custom converter functions that operate on the XML element objects may need to take this into account (for example, custom XPath querying).

23.12.2. Distributed Runtime Version Checks

To prevent unexpected bugs due to JAR version mismatches, GeoMesa can scan the distributed classpath to verify compatible versions on the distributed classpath. This behavior may be enabled by setting the system property geomesa.distributed.version.check=true.

23.12.3. Shapefile Ingestion

Shapefile ingestion through the GeoMesa command-line tools has changed to use a converter definition. This allows for on-the-fly modifications to the shapefile during ingestion, however the command now requires user confirmation. The previous behavior can be simulated by passing --force to the ingest command.

23.12.4. Delimited Text Auto-Ingestion

GeoMesa previously supported auto ingest of specially formatted delimited CSV and TSV files. This functionality has been replaced with standard ingest type inference, which works similarly but may create different results. Generally, the previous behavior can be replicated by using type inference to create a converter definition, then modifying the converter to set the feature ID to the first column ($1).

23.12.5. FileSystem Storage API Changes

The FileSystem Storage API is still considered beta-level software, and has been updated in this release. The DataStore API has not changed, however the internal class interfaces have changed in this release, potentially requiring changes in user code.

In addition, the format used to store metadata files has been updated, so older versions of GeoMesa will not be able to read metadata created with this version. When accessing older metadata for the first time, GeoMesa will update the files to the new format, potentially breaking any old clients still being used.

Finally, the update-metadata tools command has been replaced with manage-metadata.

23.12.6. Spark Version Update

GeoMesa now builds against Spark 2.3.1, and supports versions 2.2.x and 2.3.x.

23.12.7. Arrow Version Update

The version of Apache Arrow used for Arrow-encoded results has been updated from 0.6.0 to 0.10.0. Due to changes in the Arrow inter-process communication (IPC) format, clients may need to update to the same Arrow version.

23.12.8. Scalatra Version Update

The version of scalatra used for web servlets has been updated to 2.6.3. The new version requires json4s 3.5.4, which may require changes to the web server used to deploy the servlets.

23.13. Version 2.0.0 Upgrade Guide

23.13.1. GeoTools 18 and GeoServer 2.12

GeoMesa 2.0.0 is compiled against GeoTools 18.0 and GeoServer 2.12. When upgrading GeoServer instances, it’s usually best to start over with a new GeoServer data directory. If you upgrade GeoMesa in an existing GeoServer instance that has run GeoMesa 1.3.x or earlier, layers will still work but you will not be able to edit any existing GeoMesa stores. In order to edit stores, you will need to delete them and re-create them through the GeoServer UI. Alternatively, you may edit the GeoServer datastore.xml files (located in the GeoServer data directory) to match the new GeoMesa data store parameters (described below). In particular, you will need to add a namespace parameter that matches the workspace of the GeoServer store.

23.13.2. Data Store Parameters

The data store parameters used in calls to DataStoreFinder and the Spark SpatialRDDProvider have been standardized. New parameters are outlined in the individual data store pages:

The older parameter names will continue to work, but are deprecated and may be removed in future versions.

23.13.3. Removal of Joda Time

With the introduction of java.time in Java 8, the Joda Time project has been deprecated. As such, GeoMesa has removed its Joda dependency in favor of java.time. One consequence of this is that custom date patterns in geomesa-convert are interpreted slightly differently. See DateTimeFormatter for details.

Warning

In particular, “year of era” has changed from Y to y. Y now means “week-based year”, and will give different results.

23.13.4. Saxon XML Parser

The GeoMesa converter XML module now ships with Saxon-HE by default. Saxon-HE is generally much faster at parsing XML than the default Java implementation. Previously, Saxon was available as an additional download.

Warning

Saxon parsing has some differences from the default Java implementation, which may cause existing converter definitions to fail. In particular, Saxon is much stricter with XML namespaces. See Handling Namespaces with Saxon for more information.

23.13.5. Kafka Data Store

The Kafka Data Store has been rewritten into a single implementation for all supported Kafka versions. Support for Kafka 0.8 has been removed. See Kafka Data Store for more information.

23.13.6. Accumulo Standardization

In order to standardize behavior between data store implementations, some behaviors of the AccumuloDataStore have been modified.

23.13.6.1. Attribute Index Coverage

Accumulo attribute indices specified with index=true will now create full attribute indices, instead of join indices. To create a join index, explicitly specify index=join. Existing schemas are not affected.

23.13.6.2. Record Index Identifier

The Accumulo record index has been renamed to the id index. In general practice, this will have no effect, however when specifying geomesa.indices.enabled, the value id must be used in place of records.

23.13.6.3. Tools Command Name

The Accumulo command line tools script has been renamed from geomesa to geomesa-accumulo.

23.13.7. Table Splitters

The table splitting API has changed. Any custom table splitters implementing org.locationtech.geomesa.index.conf.TableSplitter will need to be updated for the new method signatures. In addition, the provided GeoMesa splitters have been deprecated and replaced. See Configuring Index Splits for more details.

23.13.8. System Properties

Time-related system properties have been standardized to all use readable durations. Durations can be specified as a number followed by a time unit, e.g. 10 minutes or 30 seconds. The following properties have been changed to accept durations, and some have been renamed. Note that this will affect system properties set in the JVM as well as any custom geomesa-site.xml files. More details can be found under Runtime Configuration or the appropriate data store configuration section.

Property

Previous name

geomesa.query.timeout

geomesa.query.timeout.millis

geomesa.metadata.expiry

N/A

geomesa.batchwriter.latency

geomesa.batchwriter.latency.millis

geomesa.batchwriter.latency

geomesa.batchwriter.latency.millis

geomesa.stats.compact.interval

geomesa.stats.compact.millis

geomesa.cassandra.read.timeout

geomesa.cassandra.read.timeout.millis

geomesa.cassandra.connection.timeout

geomesa.cassandra.connection.timeout.millis