Hi Thomas,
I’m in a similar place as you are.

I have some historical things to deal with too, which I want to solve as part of my migration/update:

• some of our tables use MyISAM (table level locking), but most use InnoDB (row level locking). I want to standardise this to InnoDB

• increasingly, records from our CRIS systems contain characters that the old default MySQL utf-8 encoding didn’t handle, as it was a 3-byte representation. Changing this to utf8_mb4 isn’t as simple as just changing the charset (see below).
• access table… it’s too big. 209 million rows. I’m going to archive it into chunks, and move it out of the database before migrating/upgrading.

Does anyone have any advice on the above?

When I’ve worked out what I’m doing, I’ll share it here/create a wiki page for it.

For the utf8_mb4 issue, these are related for new versions of EPrints:
https://github.com/eprints/eprints3.4/issues/330 (v3.4)

https://github.com/eprints/eprints3.5/issues/1 (v3.5)

but they don’t deal with upgrading a specific instance of an EPrints database.

Cheers,
John