Workflow for moving existing Portal's content directory to S3 bucket?

[JonSwoveland]

I have an existing single-machine deployment of ArcGIS Enterprise on an AWS EC2 machine, and I'm trying to move the Portal component to a new machine.  The Portal's content directory is the default local path (C:\arcgisportal\content).

This is a production system with loads of content used throughout the organization. Rebuilding it is simply not an option.

Esri has explained my only option for moving Portal is to use the join-site-and-migrate method.  This method  requires moving the existing content directory to a location accessible to the old and new Portal.  So far my attempts to setup and use a shared folder have failed miserably. 

I know it's possible to configure a new instance of Portal to use an S3 bucket for the content directory, but I have not found any documentation regarding moving an existing Portal's content to an S3 bucket to facilitate a highly available Portal deployment.   I already went through the steps of configuring an S3 bucket and copied the content directory up to it, but then discovered that when configuring the Portal to use the S3 bucket, one just specifies the bucket name, and not a full path. Therefore, if this is even possible, I would have to know how to recreate the folder structure Portal wants in the S3 bucket. 

I really hope this is possible. 

[JakeSkinner]

Hi @JonSwoveland, have you performed the steps in the following document.  See the section on the Amazon Web Services.

[JonSwoveland]

Hi Jake, thanks for your response.  In preparation for this move, I figured I'd test against a development instance of Portal I have setup.  I created the S3 bucket and copied the content directory to the bucket, using the path <bucket-name>\arcgisportal\content.   It was at precisely the step you referenced where I got stuck, because the configuration JSON contains only the bucket name, not the path.  
I have not found any documentation clarifying what the path to the content directory needs to be.  I'm guessing it is simply <bucket-name>\content,  but I've already had to restore a server from backup due to an inaccessible Portal content directory setting, so I'd like to be sure.

[JonSwoveland]

Yeah, so I just tried setting the content directory to my S3 bucket. I copied the sample JSON from the document you provided, substituted in my keys, region and bucket name, and immediately got an error: 

Error

Failed to edit portal directory 'content'. JsonObject

Code: 500

This is the snippet of JSON I used (with the keys and bucket name replaced).  I should clarify that I did not specify the bucket's S3 ARN, or S3 URI, it is simply the name with no prefix.

{
"type": "cloudStore",
"provider": "Amazon",
"connectionString": {"accessKeyId":"<access key>","secretAccessKey": "<secret key>","region": "ca-central-1","credentialType": "accessKey"},
"objectStore": "<the name of my bucket with no prefix>",
"isConnectionStringEncrypted":false
}  

Nothing in the Portal logs. 

Sigh.

[JonSwoveland · Accepted solution]

Just had a call with a very helpful engineer on Esri's advanced technical support team. She explained that the content directory structure in S3 is "flattened" so simply copying the content folder as-is to an S3 bucket won't work.
She also explained that while it is technically possible to do this, a better approach to achieve my end goal is to use WebGISDR.  I was unaware that WebGISDR could be used to restore a single machine deployment to a multi-machine deployment, but apparently it can be done. That, however, is another subject. 

[BillFox]

your move to ha will involve a lot more than that step

if you have the option, build it from scratch