Hi everyone,
I’m curious to hear your thoughts on the new Immich dataset layout, where the previously required 7 mount points (Backups, Videos, Library, Profile, Thumbs, Uploads, PGData, and MLData) are now consolidated into just 3 suggested datasets: Userdata, PGData, and MLData.
Personally, I’m not fully convinced by this change.
In my current setup, I have two separate datasets:
- One SSD-based pool for apps and ix-volumes
- One large HDD-based pool for long-term storage, backups, and important user data.
With the previous layout, I was able to configure it so that only meaningful content (original photos, videos, and database data) ends up in the large HDD dataset, while thumbnails and transcoded low-res videos—which I don’t mind losing—were stored on the SSD pool. This allowed me to save precious space and avoid backing up unnecessary files.
With the new structure, it seems more difficult to maintain that separation. All user-generated content, including low-res previews and thumbs, ends up in Userdata, making it harder to avoid backing up junk or using valuable storage for ephemeral data.
What do you all think?
Have you migrated already? Any issues or workarounds to maintain a clean separation?
Also, does anyone know when or if the old configuration will be officially deprecated?
Thanks!