How to Bulk Delete Images Based on Name Pattern in Immich?

[coltenkrauter]

I’m in a bit of a bind and could use some guidance on cleaning up my Immich server.

I need to delete approximately 500,000 images from my Immich server that follow the naming pattern *-thumbnail.webp. These files were autogenerated thumbnails at some point, but I accidentally re-uploaded them as part of my library due to a deployment mishap.

Backstory

• I originally deployed Immich using the TrueCharts app on TrueNAS.
• Due to TrueNAS Scale moving away from TrueCharts (TrueNAS migration discussion), I set up a new Immich instance using the TrueNAS Scale app.
• During the migration, I pointed the Immich CLI to my photos directory, which also contained the generated thumbnails. As a result, both my library and the thumbnails were uploaded, leading to the month of May 2024 containing a low-resolution copy of my entire library.
• Later, I switched to Docker Compose for a more stable setup. This time, I was more careful when uploading to not include the thumbnails dir... but I had not yet noticed that I already had thousands of thumbnail images mixed into my library.

Issue

I would like to delete all *-thumbnail.webp files, but I want to ensure that I clear both the files and any associated thumbnails and entries in the Immich database to avoid orphaned references or issues.

I understand that I could theoretically run something like this,

find /path/to/root -type f -name '*-thumbnail.webp' -delete

However, I suspect this wouldn’t clean the Immich database properly, and I'm not sure if there would be a bunch of thumbnails of thumbnails wasting space on my server.

Question

Is there a recommended or safe way to bulk delete these files and also update the Immich database?
Or am I better off taking the "nuclear option" of setting up a fresh instance and re-uploading only the correct files?

The latter would be time-consuming as my library is nearly 5TB in size and would take many days to re-upload, so I’m hoping there’s a more efficient solution.

Any advice or guidance would be greatly appreciated!

[bo0tzz]

From the advanced search menu, you can filter by (part of) filename:

Then you could try just select-all and delete, though for half a million assets the UI might not be all too happy about that. A better chance would be to write a little script using that search query against the API directly.

[coltenkrauter]

From the advanced search menu, you can filter by (part of) filename:

Then you could try just select-all and delete, though for half a million assets the UI might not be all too happy about that. A better chance would be to write a little script using that search query against the API directly.

Great idea, the API. 🤦
That's what I'll do! Thank you!

I had seen the UI option, but I was concerned that it might not be able to handle the scale.

[coltenkrauter]

I whipped up this script to bulk delete for me.

[coltenkrauter]

Worked great... I cleared cache in my browser... Now, the timeline seems to have some gray placeholder box for files that got deleted...

Just checked the logs, it is showing many of these,

Queueing 1000 assets for deletion from the trash

So maybe deletions are still in progress?

[coltenkrauter]

Okay, found out that I also need to blow away all of the *-preview.webp files.... So, that will probably fix me up.

[coltenkrauter]

Yup... Things are good now.

[coltenkrauter]

I will note that I experienced something like this while all of these thumbnails were uploaded on the same day: #13319