To use GeoAnalytics Server you need a GeoAnalytics Server license. This is a separate license that's applied to ArcGIS Server. After you apply this license, federate your server with the ArcGIS Enterprise portal, and set your server as your GeoAnalytics Server in your portal under Organization > Settings > Servers.

Learn more about setting up GeoAnalytics Server

Your GeoAnalytics Server site can contain one or three machines. These machines are in addition to your base deployment and your machines that have ArcGIS Data Store configured as a spatiotemporal data store.

Learn more about best practices for GeoAnalytics Server sites

Federate GeoAnalytics Server on the Organization > Settings > Servers page in your ArcGIS Enterprise portal.

Learn more about federating GeoAnalytics Server

Register ArcGIS Data Store (relational, spatiotemporal big data store, and tile cache) with your hosting server.

Yes. GeoAnalytics Server works with the same spatiotemporal big data store as ArcGIS GeoEvent Server.

Other than the license applied, creating a multiple-machine site for GeoAnalytics Server is no different than setting up a multiple-machine site for ArcGIS Server. Ensure that you have a shared configuration store that's available to all server machines and that all machines have the same specifications (number of cores and amount of RAM). If you're scaling from a one-machine site to a three-machine site, make sure your current configuration store is accessible to all machines.

An administrator sets the percentage of your machine resources (cores and memory) through the GeoAnalytics Server Administrator Directory system properties setting. You can check the amount using the Spark health check.

Administrators should reset Zookeeper in the following circumstances:

• After resetting or changing the service properties (the percentage of cores and RAM used for GeoAnalytics Server)
• If Zookeeper isn’t running (after adding or removing machines, for example) or if distributed compute errors occur after analysis

Learn more about resetting Zookeeper

You can use data that you connect to as big data files shares, as well as layers in your portal.

You can connect to the following sources as big data file shares:

• Apache Hadoop Distributed File System (HDFS), with or without Kerberos authentication.
• Apache Hive
• Folder and network shares
• Cloud Stores—With support for the following:
  • Amazon S3
  • Microsoft Azure Data Lake
  • Microsoft Azure Blob storage

You can use the following layers as input:

• Hosted feature layers (including views) stored in the ArcGIS Data Store (configured as spatiotemporal big data store or relational data store)
• Feature services
• Stream services (this applies analysis on data currently exposed in the layer)

You can write to hosted layers in your portal, or use a template to write to using a big data file share.

When you write to hosted layers in your portal, specify whether you want to save to the spatiotemporal data store (the default) or a relational data store.

You can write to the following data stores by configuring a big data file share template:

• HDFS
• Folder and network shares
• Cloud Stores—With support for the following:
  • Amazon S3
  • Microsoft Azure Data Lake

You register a big data file share using ArcGIS Server Manager on the GeoAnalytics Server. In Manager, click Site > Data Stores > Register a big data file share. You can also register a big data file share using ArcGIS API for Python.

Confirm the following to ensure that you can register a big data file share:

• The server is configured as the GeoAnalytics server of your portal.
• You're using Manager on the GeoAnalytics Server.
• You're a publisher, administrator, or user with the GeoAnalytics Feature Analysis privilege.

Make sure your GeoAnalytics Server Run As account has been given read access to the location you are trying to register. To learn about registering a big data file share, see Register your data with ArcGIS Server using Manager.

If a manifest contains no datasets, it may be due to one of the following:

• The location selected for the big data file share did not contain the correct folder structure. To learn about the correct folder structure, see Register your data with ArcGIS Server using Manager.
• The datasets are not in the required format. To learn more about the types of data you can register using a big data file share, see Get started with big data file shares.
• If you're registering a cloud store as a big data file share, make sure you correctly specified the folder you're using when registering the cloud store. Cloud store validation occurs even if the folder isn't correctly specified.

If you're still unable to register datasets as part of your big data file share, contact Esri Support, and they can assist you in registering your big data file share.

A dataset may be skipped in manifest generation if there are unexpected characters, it's an empty file (or empty lines), or the file is corrupt. If you're trying to register a delimited file, open the file and check for extra characters, such as unescaped delimiters on each line. You only need to preview the first 50 lines in the dataset, as manifest generation is based on a sample. If your dataset is composed of shapefiles, verify that all the shapefiles have the same schema, that none are greater than 2 GB in size, and that they aren't corrupt. You can test for corruption by opening them in ArcGIS Pro. If you're still unable to register a dataset, contact Esri Support.

When you register a delimited file as a dataset in a big data file share, GeoAnalytics Server tries to determine the field names. Depending on your data, you may not have field names, or GeoAnalytics Server couldn't determine them. In this case, you have the following options:

• If your dataset doesn't have headers in the original data, you can create headers by setting the fieldNames parameter in the hints file, uploading the hints file, and regenerating your manifest in Server Manager.
• If the manifest generation incorrectly presumed the delimiter type of your data and couldn't parse the fields, set a delimiter parameter in the hints file, upload the hints file, and regenerate your manifest in Server Manager.
• If your data has a header row and the delimiter was incorrectly discovered, set the hasHeaderRow parameter in the hints file, upload the hints file, and regenerate your manifest in Server Manager.
• Manually modify the field names in the manifest editor.

By default, a big data file share is accessible only to administrators and the publisher who registered the data store. A portal item is created in your My Content directory of the portal when a big data file share is registered. Share this portal item with the other publishers.

No. GeoAnalytics Server will run analysis on the data source that's registered, even when records are added, modified, or removed. You do need to modify the manifest if the schema of the data changes.

Do a visual inspection of your dataset fields in Server Manager. If they look okay, use the input in the Describe Dataset tool. Using this tool, you can create a sample of your input that has the same time and geometry as your inputs, as well as an extent layer representing the spatial extent of your data, summary statistics on the input fields, and a description of the registered time and geometry.

Yes; although there are a few considerations.

If your data is exposed as a feature service, you can use that as input to your tool. However, you'll experience better performance if you create a hosted feature layer on your portal with GeoAnalytics Server. This is because the GeoAnalytics Server read the data directly from the data source. If your data is not stored in a feature service, you have the following options:

• If you're running the tool in ArcGIS Pro and your data is small and not complex (such as a few hundred points), you can use your data as input to the tool directly in ArcGIS Pro. In most cases, it is recommended that you share your data as a hosted feature layer first.
• When you use data stored in a file geodatabase, your data must be copied to your GeoAnalytics Server before you can complete analysis. This can take a long time, and in some cases fail to complete, depending on the size of your data.

To determine whether a GeoAnalytics Server job is running, do any of the following:

• In Server Manager on your GeoAnalytics Server, view jobs using the Jobs tab.
• View the jobs in your GeoAnalytics Server Administrator Directory by clicking admin > services > System > GeoAnalyticsTools.GPServer > jobs.
• View the activeapps in the GeoAnalytics Server health check.

Because of the way distributed processing is implemented, it's hard to tell how long a job will take. Each job is created on tasks that are composed of one or more stages. The number of tasks and stages will differ depending on the number of cores, the tool, and the number of inputs you're using. If you're running a tool in ArcGIS REST API or ArcGIS Pro, updates on a task's progression are provided. Each task is a different size; some may take seconds, and some may take hours.

In addition to clicking Cancel in ArcGIS Pro or Map Viewer, you can cancel a GeoAnalytics Server job using either of the following options:

• In Server Manager on your GeoAnalytics Server, use the Jobs tab.
• In the GeoAnalytics Server Administrator Directory, click admin > services > System > GeoAnalyticsTools.GPServer > jobs, query for running jobs, and cancel the job.

You can tell whether your GeoAnalytics Server job has started in ArcGIS Pro by looking for the following text in the geoprocessing messages: Executing (<Tool Name>): <Tool Name> <input parameters>. If you don't see that text, your GeoAnalytics Server job hasn't started, and the data is still being copied to the GeoAnalytics Server. To avoid waiting for data to copy, use hosted feature layers or big data file shares as input.

Run another tool to see if the same error occurs. In most cases, your best option is to contact Esri Support so they can help isolate what's causing the issue. It may be related to the parameters you're using, the input layers, or a distributed compute platform service error. Your administrator can reset Zookeeper if you think it may be a compute platform service error. When contacting Esri Support, note the parameters you're using and whether you've run other GeoAnalytics Server tools successfully.

The job you're running requires more resources than you have available on your GeoAnalytics Server. The cause could be any of the following:

• You're already running a GeoAnalytics Tools on your GeoAnalytics Server, so the resources are not available for a new job. If this is the case, cancel the current job or wait for it to complete before starting the new job.
• You may be using incorrect settings for the amount of resources. To see how much RAM and the number of cores that are available on your machine and being requested for a job, use the GeoAnalytics Server settings.
• There may be an issue with Spark. If you've ruled out the above scenarios, try restarting the GeoAnalytics Tools through Server Manager.

To fix this, restart your GeoAnalytics Server using the following steps:

1. Stop the ArcGIS Server service on all machines that are part of your GeoAnalytics Server site. On Windows, stop the ArcGIS Server Windows services. On Linux, stop all machines using the stopserver.sh script.
2. Wait five minutes.
3. Start the ArcGIS Server service on all machines in your GeoAnalytics Server site. On Windows, start the ArcGIS Server Windows service. On Linux, start all machines using the startserver.sh script.
4. Wait two minutes. Then run a GeoAnalytics job. If it doesn't work, continue to step 5.
5. Check the status of the compute platform service. To do this, see the steps below.

The compute platform service can be started in the ArcGIS Server Administrator Directory on the GeoAnalytics Server. Complete the following steps to restart your compute platform service:

1. Sign in as administrator to the ArcGIS Server Administrator Directory of your GeoAnalytics Server.
2. Go to system > platformservices > Compute_Platform > stop and click the stop button.
3. Go to system > platformservices > synchronization_service > stop and click the stop button.
4. Wait five minutes.
5. Go to systemplatformservices > synchronization_service > start and click the start button.
6. Go to system > platformservices > Compute_Platform > start and click the start button.
7. Wait two minutes for the machines to start.
8. Go to system > platformservices > Compute_Platform > status, and check that all the machines have a configuredState and realtimeState condition of started.

If the platform service does not start, contact Esri Support.

If your job immediately fails with the above error and doesn't return any more information, your job wasn't correctly started. Restart your GeoAnalytics Server in Server Manager on your GeoAnalytics Server machine.

If you ran this job in ArcGIS REST API or ArcGIS Pro, you may have received an additional message Processing request took longer than the usage timeout for service 'System/GeoAnalyticsTools.GPServer'. This error occurs when the job you're running takes longer than the GeoAnalytics Server time-out period. To fix this, increase the time-out period. See GeoAnalytics settings for more information.

When the primary machine in a GeoAnalytics Server multiple-machine site becomes unavailable or is removed from the site, the site goes into recovery mode to select a new primary machine. This process typically takes six to seven minutes, and GeoAnalytics Tools cannot be run during this time. For information about the primary machine and the status of your GeoAnalytics Server machines, use the compute platform health check operation in the ArcGIS Server Administrator Directory.

Learn more about the GeoAnalytics health check operation

When you use an HTTPS URL, verify that the URL to the layer does not use a self-signed certificate.

Analysis can only be run on big data file shares that have all the required information for analysis. In ArcGIS Pro, this means that the manifest for your big data file share includes a geometry and spatial reference. If your dataset includes time, you must have the fields representing time denoted, in addition to the role. To learn more about big data file shares, see Understanding a big data file share manifest.

In some cases, this may be expected. For example, if you select a process extent with no data in it, no results will be returned. If it's unexpected, check the inputs and environment settings of the tool you used.