The recommended approach would be to restore the backup from 7.6 on a separate 7.6 instance - upgrade it all the way up to your destination version, and then create a new backup from that intermediate 8.20.11 instance - then you get a compatible zip.
With CSV, you can import attachments - https://confluence.atlassian.com/adminjiraserver/importing-data-from-csv-938847533.html
It supports http/s and file protocols.
So this is quite a job to figure out a good way to do it, but it's possible. Granted though you probably would need somebody confident with scripting to generate the .csv entries for it. Note that the url it downloads the file from in this case must be either authenticated as part of the url, or anonymously accessible. So with 7.6 this could be an issue since you cannot use any tokens in the url. In 7.6 you can still use "os_username" and "os_password" to get around the auth, although of course needless to say this is plain texted so you really only want to do this in a secure network and ideally with a discardable, temporary account. Other than that you might have to move those attachments elsewhere, could be a public apache directory, or a different server entirely. And that would also need to be done so you can clearly identify which attachments belong to which issues, so you can them use appropriate paths for each issue row.
So yes, it's possible, but not for the faint of heart. Unless resources are a constraint then spinning up an intermediate instance might be easier and would give you more complete data - such as history, comments, the things that CSV import doesn't support. The intermediate instance can be any spare machine/VM really - since you just need it to work, it doesn't have to lift any user traffic or be fast, so long as it won't run out of memory or disk space.
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.