I can't speak to the ePrints side of things, but on the Fedora side, there are nominally two options to consider:

1) create your records programmatically, via the REST API
2) create FOXML serializations of your ePrints records and then ingest those into Fedora

There are a number of variations on either approach, depending on the number/size of records involved to optimize ingest times. My general preference would be for #1.

Some best practices (not specific to migration) to consider in your Fedora repository:
i) use (M)anaged datastreams rather than Inline (X)ML datastreams where possible
ii) use Akubra and not the legacy low-level store
iii) unless you have a fairly small repository, update the HashIdPathMapper config (in akubra-llstore.xml) to have a deeper hierarchy (e.g. "#/##" or "##/##) rather than just "##".

I'd recommend joining the Fedora users mailing list for any Fedora-specific follow-on.


On May 9, 2013, at 12:16 AM, "Williamson, Kelsey CTR NUWC NWPT" <[log in to unmask]> wrote:

> Good afternoon,
> I was hoping someone might be willing to talk about migration projects. I am getting ready to migrate records from an ePrints repository into a fedora repository- but I've never done anything like this before. I have an abstract plan, but I think hearing other people's experiences would be very helpful; particularly any lessons learned?
> Best regards,
> Kelsey