PORTAL HA

[SandeepBurra4]

Dear All,

I would like to ask one question which can help me understand more on Portal HA.

We followed link for 10.5.1 to configure Portal HA Configure a highly available portal—Portal for ArcGIS (Windows) Installation Guide (10.5) |

ArcGIS Enterprise One is shown as Primary and the other is Standby. Incase the primary portal goes down, it takes about 3 minutes for the standby to become active and till then portal is not accessible. Why it says Active/Standby and not Active/Active? Is it because of software license restriction or is this how it is done?

When we checked ESRI documentation, it says Portal supports Active/Active.

Can you please more details on this.

Looking forward on this. 

[RyanUthoff]

With Portal (and ArcGIS Data Store), there can only be a primary and standby machine. The primary Portal machine replicates to the standby Portal machine. When the primary machine fails, the standby machine is promoted to primary.

Server on the other hand supports active/active.

This link will explain configuring a highly available ArcGIS Enterprise deployment.

Configure highly available ArcGIS Enterprise—Portal for ArcGIS (10.6) | ArcGIS Enterprise 

[JonathanQuinn]

I'll add that the web server in Portal is active/active, so both machines take requests. Only one database is active, though, on the primary, so eventually all requests are directed to the primary.

The limitation is mostly on the database. The database, (postgres), only supports active/passive out of the box.

At 10.6.1, the failover time should be under a minute, typically 30 seconds or less. Prior to that, the failover time you see is expected.

[MathiasGieser]

I am having some strange issues that I cannot yet pinpoint to our own network infrastructure or something in the configuration.  Luckily this is only a test environment for us and its the perfect opportunity for these sort of things.

We are using a third party load balancer for our two portal servers, specifically it is a Netscaler.  We have the health check urls in place as well.  The documentation and setup led us to believe it would be active-active setup.  However, after speaking with support on another case and reading through the forums I now understand it isn't truly active-active.

You mention in your comment that both servers still handle requests.  With that statement in mind, is it still acceptable to send requests to either server?

Thanks for the time and your response.

Mathias Gieser
mgieser@bepc.com

[JonathanQuinn]

The web servers on each machine are both active and capable of receiving requests, so in that sense, it's active-active. There is a primary database on the primary Portal, though. The standby will connect to the database on the primary machine when it receives a request that requires data from the database, so it's active/passive at the DB tier.

There's no problem sending requests to each machine.

[MathiasGieser]

Thank you for clarifying this.  Great explanation!

[VHolubec]

Dear Jonathan,

in the other words is it right that the primary purpose of HA Portal for ArcGIS is to eliminate the single point of failure not to scale the performance of Portal for ArcGIS as it in case of ArcGIS Server where adding additional machines increase the performance of ArcGIS Server? 

Thank you for reply. 

[JonathanQuinn]

Right, setting up an HA Portal isn't necessarily for more throughput, as the Portal is essentially just a website and all requests are making it into the primary Portal machine eventually. It's more so for redundancy/eliminating a single point of failure. And yes, setting up a multi-machine Server site serves two purposes, more processing power but also adding redundancy.

[VHolubec]

Dear Jonathan,

thank you very much!.

[CameronBlandy]

Hi,

Jumping into this thread. We discovered a major issue with our HA Portal environment. If we turn the primary machine off the portal switches over to the other machine as expected and makes it primary. Problem is, that the load balancer still tries to direct traffic to the now off machine causing portal to hang as it knows no way to send traffic to the primary machine.

Has anybody else encountered this issue? How do we configure the load balancer to only direct traffic to the primary if the standby is actually switched off?

With all the tests we have been doing (switching machines off and on) we seem to be in error state where portal does not know which machine is the primary. We can no longer log in to portal/home or portaladmin. Checking the logs shows an error: HA: Master down, check if this node needs to be elevated to be the new master.

Given that we cannot log in how can we change the Master to the machine that is currently switched on? We are using 10.6.1.

Cameron