ArcGIS Server 10.7.1: Does the High Availability (HA) of VM environment replaces clustering and failover techniques?

[JamalNUMAN]

ArcGIS Server 10.7.1: Does the High Availability (HA) of VM environment replaces clustering and failover techniques?

I couldn’t figure out if there is still a need to go for clustering or failover techniques in case all GIS servers are based on VM and High Availability (HA) is ensured

[DanPatterson]

Are you following this?

Steps to get ArcGIS Server up and running—ArcGIS Server (Windows) Installation Guide | Documentation... 

Why not contact your regional office to get help and confirmation to your questions.  They would most likely be able to support better with reference to your specific locale

[JamalNUMAN]

Thanks Dan.

The Geonet is better place as experts share their experience based on concrete practice.

I went through the available documents but never found a precise answer.

[JoshuaBixby]

A VM HA environment reduces the chances of having hardware related outages, but it doesn't change the fact that operating systems and server applications can get corrupted or otherwise go offline.  Since ArcGIS Enterprise isn't containerized, yet, spinning up a whole new VM and installing and configuring ArcGIS Enterprise might take some time, unless you have a staging server you can fall back to.  And, the most important question of all, how much downtime are you willing to live with vs the cost of running more VMs and licensing more ArcGIS software.

It depends on so many user or environment specific factors that it is impossible to make a blanket statement.

[JamalNUMAN]

Thank you Joshua for the useful input.

I thought that the VM HV is a perfect solution as any VM fails it is immediately replaced with an equivalent one with same content in negligible time.

[JoshuaBixby]

Although I am not sure your definition of "negligible time," you appear to have a lot of faith in your VM HA physical environment and operating procedures.  Different VM HA environments are set up and operate differently.  Typically the failover at the hardware level is seamless, or nearly seamless, but recovery at the VM level is typically not seamless.  Have you done any disaster recovery exercises or had a machine/application go down (not due to hardware/infrastructure issues) to test how recovery goes?

[JamalNUMAN]

Thanks Joshua

I’m referring here to case#4 in the configurations below. Suppose that a physical machine is divided into 3 VMs. Then in case one of these VMs is down, then a new machine spins up to replace the corrupted one.

[JoshuaBixby]

Where did that documentation screenshot come from?

Does ArcGIS Enterprise support Single-machine high-availability (active-active) deployment—ArcGIS Enterprise | Documentation for Ar... ?  Yes.  Is HA active-active a replacement for multi-machine/clustering and failover?  Possibly.

There is no single method of deploying high-availability with ArcGIS Enterprise.  Each method has trade-offs, and which method is best involves balancing those trade-offs. 

[JamalNUMAN]

Thanks Joshua

The reference of the screenshot is here:

https://wso2.com/whitepapers/capacity-planning-for-application-design-part-1/#02

Your input is very useful and provides clear idea in how things are managed