We have a load balanced Episerver 9 Commerce solution on Amazon, hosted on individual EC2 servers, 2 frontend instances and 1 authoring instance. Authoring instance is generating the full text indexes and save it to the file system which is shared between the 2 frontend instances.Would it be possible to save the full text indexes to an S3 bucket and the frontend instances read from there instead on relying on a shared folder on the filesystem?
Is there any example I could use to implement this behaviour? Is it advisable to do it this way?
This is not supported for Amazon. You might want to consider using Episerver Search & Navigation (formerly Episerver Find).
Or, if you are up for a challenge, you can build a provider that emulates how the LuceneAzureSearchProvider does what you need (although it's made for Azure).
The default LuceneSearchProvider relies on the file system, so that does not work. Implementation of LuceneAzureSearchProvider is supported by https://github.com/azure-contrib/AzureDirectory which greatly simplifies reading/writing to blob storage
It seems you can try this out, but there is no guarantee that it'll work