Maintenance

Cleanup

When searching for tags and other subset filters, marv presents lists of existing tags. To present only used items, these lists need to be regularly cleaned-up.

marv cleanup --unused-tags

When datasets are deleted (via frontend, API, cli) they are only marked as discarded and can be “undiscarded”.

marv undiscard --help

To actually delete these

marv cleanup --discarded

Additional cleanup operations will be added in one of the next releases.

Backup

By default, all data related to a marv site is stored in a site directory. It holds the following information:

  • db marv’s sqlite database
  • marv.conf the marv configuration file
  • sessionkey if it changes, users will have to relogin
  • store of all node output, theoretically possible to recreate by running all nodes, which might not be feasible
  • uwsgi.conf configuration for uwsgi serving the marv application

Make sure to create regular backups of this site directory or the individual components in case you placed them elsewhere. It is not necessary to stop marv to create backups.

Dump/restore

Use marv dump and marv restore to dump marv’s database to a json format and to restore it in a site with the same configuration. You can use the latest version to dump older databases.

The dump contains:

  • datasets for all collections with setid, files, tags, and comments
  • and users with group memberships.

Apart from the dump, you need to copy from old site:

  • marv.conf
  • uwsgi.conf
  • sessionkey (if it changes, users will have to relogin)
  • store (if you don’t copy the store, all nodes will have to rerun)
cd old-site
marv dump ../dump.json

cd ../new-site
cp -a ../old-site/marv.conf ./
cp -a ../old-site/uwsgi.conf ./
cp -a ../old-site/sessionkey ./
cp -a ../old-site/store ./
marv restore ../dump.json

DUMP/RESTORE IS NOT A REPLACEMENT FOR MAKING BACKUPS