Slow performance after a few uses

[Kobelayton]

Awesome work. It worked amazing on my first couple of takeouts. I didn't do the wildcard for all of them because I wasn't sure how it was going to work. After doing a few 2gb takeouts, the speed went to a crawl. Like 1 asset every 10s. I even deleted previous uploads so there was no conflict with previous versions. Restarted the PC and my Synology; still a crawl.

Thank you for the great work, I'm not sure what I'm missing.

[simulot]

Immich-go itself need some time at startup to build file indexes. Once done, it just feed immich with images.
So I guess that the slowness is due to the immich server. And a good reason of the slowness is the power of the server.

Note that movie files are longer to process on the server side.

[Kobelayton]

No, this is the upload phase that's slow. Again, I did a few 2gb takeout files to test the waters and they were lightning fast. Started getting progressively slower. Now it's a crawl.

Edit: I did have a remote server doing the machine learning and I thought that might be it, but sadly no.

[simulot]

Immich-go just passes the file to the server. It depends on the server speed, except for the file indexing at startup.

[Kobelayton]

Thanks, I'll have to keep investigating. I've confirmed full Gb transfer rates via Libre and transferred a folder tree with thousands of images and it goes very fast. For some reason I'm still very sluggish on immich-go; approximately 8s per image.

[simulot]

When I run immich on my machine, 2% of the CPU immich-go, the rest ofor immic (gunicorn, immich, ffmpeg, perl...)

Once again, check your configuration.

[Kobelayton]

Similar, I don't think it's an issue with go. I think I have something going on w/ my NAS. Container manager is slow and unresponsive, multiple restarts etc.

[Kobelayton]

Follow up - it seems closing all other docker containers running has boosted my performance considerably. It's now 1-3s per image.

[simulot]

Better, but still not fast. You can allocate more ram to immich to see if it helps