A web GIS is comprised of the following:
- Web adaptors* and at least one load balancer
- A domain name system (optional)
- A file server for the ArcGIS Server configuration store
- One or more GIS servers (ArcGIS Server sites*)
- Data stores (including registered databases, files in registered folders, and a managed database or ArcGIS Data Store*)
- A portal (in on-premises deployments, this is Portal for ArcGIS*)
ArcGIS offers tools and functionality that allow you to configure high availability for those components listed with an asterisk (*). The other components require use of third-party tools and functionality to configure.
Highly available GIS servers
Web GIS deployments contain a hosting server. This is a GIS server you have dedicated to running your portal's hosted services, such as hosted feature, tile, and scene layers. You might choose to federate additional GIS servers with your portal to allow them to share authentication and to automatically register web services as items in your portal. Or you can add services from stand-alone GIS servers as items to your portal. You can configure any of these GIS servers to be highly available.
How you implement a highly available GIS server depends on whether your web services reside on a single machine or are spread across multiple machines.
If your services run on one machine, configure a primary and a standby site. Also configure a network load balancer to communicate with the active GIS server machine. The primary and standby sites will each have their own server directories and configuration store. These directories must be identical.
If your GIS server is static (for example, you have published the web services you need and won't be publishing more), you can configure both the primary and standby sites to be active. In this case, the primary and standby share an output directory that you configure on a highly available file server. However, this is an unlikely scenario when your GIS server is used with Portal for ArcGIS and most likely would not be used.
If you have multiple machines in your site, configure a network load balancer to communicate with your pool of GIS servers. This pool of GIS servers shares server directories and a configuration store. Again, configure these directories on a highly available file server to implement a highly available GIS server.
Highly available data stores
Use ArcGIS Data Store as the data store for your portal's hosting server and you can take advantage of the failover functionality it provides.
For highly available hosted feature layer data, install ArcGIS Data Store on two separate machines and create a relational data store on each machine. Configure each relational data store with the GIS server you will use with your portal's hosting server. The first relational data store you configure is the primary relational data store machine; the second machine you configure is the standby.
For highly available hosted scene layer caches, install ArcGIS Data Store on two separate machines and create a tile cache data store on each machine. Configure each tile cache data store with the GIS server you will use with your portal's hosting server. The first tile cache data store you configure is the primary tile cache data store machine; the second machine you configure is the standby.
ArcGIS Data Store automatically replicates hosted feature layer data and hosted scene layer caches from the primary data store to the standby; therefore, the data exists in two places. The GIS server will always communicate with the active (primary) data store.
For highly available archived observation data used with ArcGIS GeoEvent Extension for Server, highly available, you can install ArcGIS Data Store on three or more machines and create a spatiotemporal big data store on each. Configure each one with your portal's hosting server. A copy of each data set exists on at least two of the data store machines at any time. If one machine fails, the data store ensures that at least two of the remaining machines contain the data.
Highly available portals
A highly available portal includes two Portal for ArcGIS machines accessed through a network load balancer. The load balancer redirects incoming requests to the active portal machine.
The two portal machines store content in a common directory. For your portal to be highly available, you must configure this content directory on a highly available file server.
Combining high availability and disaster recovery
In most cases, you will implement a disaster recovery plan in addition to a highly available deployment. That way, if all the machines in your deployment are lost (such as in a natural disaster), you still have a backup of your data and services you can use to bring your deployment back online. Either maintain backups in a secure, offsite location or maintain a disconnected standby deployment in a remote location.