Select to view content in your preferred language

Configuring the ArcGIS Data Store in an Active/Passive HA Implementation

1067
2
06-30-2022 07:04 AM
AndrewThePrivateSectorDrone
New Contributor II

A user posted a similar question a couple years ago, but that post wasn't answered directly so I figured I would revive the topic. 

 

The environment I am working in extends the Single-Machine Active-Passive HA implementation pattern. The simplest representation of this deployment is displayed below:

Link to Architecture Image 

I use the word 'extend' because, in addition to the ArcGIS Server, this deployment also leverages Portal and the ArcGIS Data Store. So in my environment, we have two (2) separate deployments of ArcGIS Enterprise (AGE) that are positioned behind a network load balancer (NLB) that routes traffic to the appropriate system. Under normal conditions, the NLB routes traffic entirely to the Active AGE; when the Active AGE fails a health-check, the NLB routes traffic to the passive AGE. We have a scripted process in place to keep the ArcGIS Server services/resources in sync between the two (2) deployments.

 

My question is aimed at the ArcGIS Data Store in this environment. Specifically, my question is: how is the ArcGIS Data Store supposed to be configured in an Active-Passive deployment of AGE? Logically, it would make sense that each Data Store be registered with the ArcGIS Server participating in the node (i.e. Active DS --> Active Server; Passive DS --> Passive Server), but this introduces the complication related to keeping independent ArcGIS Data Stores in sync between sites (which seems potentially possible through a customized use of the webgisdr utility, but I haven't tested that workflow yet). 

Another user's response to a related question suggested that the Data Stores be registered "together" and operate in Primary/Standby mode, which seems anti-pattern as that would establish a shared/common resource between the otherwise completely separated systems. 

Should the Data Stores be registered 'together' and operate as a multi-machine DS implement [and thus rely on their own failover systems]? Or should the Data Stores be registered independently and be kept in sync via the webgisdr utility or another process? Or am I missing the appropriate option?

Any assistance is much appreciated. 

0 Kudos
2 Replies
ReeseFacendini
Esri Regular Contributor

If Enterprise is set up to be highly available, with both Portal and ArcGIS Server configured in a joined site method, the passive ArcGIS DataStore can be registered with the "failover" ArcGIS Server. ArcGIS Server doesn't function in active/passive, but rather active/active. While you do have to pick a Server URL for registration, DataStore gets registered with the complete ArcGIS Server site (which would be 2 nodes in this case) and completes a sync itself while it's idle

0 Kudos
AndrewThePrivateSectorDrone
New Contributor II

Thanks for replying, Reese!

The systems are not joined and instead follow the Active/Passive HA architecture (documented here). Likewise, the ArcGIS Servers (and the entire stack) do function in an Active/Passive capacity via the NLB. 

In a scenario that extends the pattern shown here, I am interested to know if the ArcGIS Data Stores are intended to be individually registered with the appropriate ArcGIS Server Site (which would then require some form of non-joined synchronization). Does that clarify what I am asking? 

0 Kudos