Docs: improve the backup section for v2.0

[sphuber]

Fixes #4853
Fixes #2603

With AiiDA v2.0 the repository is now provided by the disk-objectstore
which is optimized for being backed up with rsync. This means the old
custom implementation of backing it up is no longer necessary nor
supported. It can now simply be backed up using rsync.

The instructions are updated and improved by tying the backup procedure
to the storage backend used for the profile. This will allow to have
other instructions for other storage backends in the future, as they
will most likely not be the same.

[sphuber]

This should be merged after the release of v2.0.0b1

[chrisjsewell]

This should be merged after the release of v2.0.0b1

Thanks but why do you say this?
You know that the current documentation is currently set to 1.6.7 as latest? https://aiida.readthedocs.io/projects/aiida-core/en/latest/
and users can select other versions of the documentation

All the documentation on the release of v2.0.0b1 should relate to that version, and I would also like to link to this in the change log.
I would recommend having an additional note admonition for "django/sqlalchemy (legacy)", or similar, just mentioning that the backup would be identical to psql_dos

[sphuber]

Thanks but why do you say this?

I added this when I realized that this is not the documentation that users should read when wanting to migrate to v2.0. As I later then realized and said in the Slack channel, we should simply add a link in the email that points to 1.6.7 instead of latest such that even when latest becomes the current develop, people will still get the correct documentation. So indeed this can be merged now.

All the documentation on the release of v2.0.0b1 should relate to that version, and I would also like to link to this in the change log. I would recommend having an additional note admonition for "django/sqlalchemy (legacy)", or similar, just mentioning that the backup would be identical to psql_dos

We could add the note for legacy backends, but it is not identical to psql_dos. At least you could use the same approach, but for large repo's that is not really feasible and so that is why there was this whole custom backup schema. This was only available in v1.x and has been removed in v2.x so it doesn't make sense to document this in v2.x. I would then maybe just in the v2.0 docs have a link to the old v1.6.x docs of how to use the old custom backup feature.

[chrisjsewell]

This was only available in v1.x and has been removed in v2.x

This feels confusing though.
If we are talking about them backing up before migrating a v1.x profile, then (a) won't they have v2.x installed, (b) if they have already used verdi then the config will have already been migrated, so actually it will point to the psql_dos backend name

[sphuber]

If we are talking about them backing up before migrating a v1.x profile, then (a) won't they have v2.x installed

Not necessarily. I guess we should in this case actually stress that if they use the old custom backup script for their backups, they should create a backup before installing v2.0.0b1, and only once the backup is complete, update their installation and environment.

(b) if they have already used verdi then the config will have already been migrated, so actually it will point to the psql_dos backend name

Yes, but they won't have run the migration yet, and so the suggested method for the psql_dos backup will be horrendously slow for mediumly sized database and everything above.

So conclusion: I think the only solution then is to make it clear in the email that people with large databases that have been using the custom backup script (if they have never used it, then there is no point, they might as well rsync it) should backup before installing v2.0. Not sure if we should then also incorporate this is in the documentation. In the backup section itself won't make much sense, because it might come too late and only read it after having updated the installation.

[chrisjsewell]

I think the only solution then is to make it clear in the email that people with large databases that have been using the custom backup script (if they have never used it, then there is no point, they might as well rsync it) should backup before installing v2.0.

I feel then that this should also be added to #5426, as the "canonical" place for how to go from v1.6 -> v2.0, or add something on the wiki and link to that?
Obviously, not all people with a v1.6 database may get the email or read it, so should there should be a more "permanent" place for such information

[chrisjsewell]

Still a question for me, of where to put the "canonical" instructions for backing up, before migrating a 1.6 database (#5431 (comment)).
But I don't think that blocks this 👍