I used API import to populate a new instance of Omeka classic using AWS S3 storage. Everything came over fine, except for the thumbnails. As a test, I tried copied one of the thumbnails from the previous server to the appropriate S3 bucket and it worked. It’s been a few months since I tested that, and now that I’m trying to scale this up that trick no longer appears to be working. (Note, the one I did copy over still displays, but the new ones do not.
Also, new items that are loaded directly into Omeka correctly create and display derivative images of all sizes.
Also, I’m using Omeka 2.61, which was updated from 2.5. I’m on Hello Roy, although switching out the themes doesn’t seem to matter. I’ve made no customization to the php files.
When when I manually copied the thumbnail from the originating server to the S3 bucket the second draws from, I figured it would work again. That thumbnail continues to display, but when I try to replicate that process now it doesn’t work.
If the original files are copied over, and it’s just the derivative images which aren’t being generated, you could try running Derivative Images and see if that creates new thumbnails for the existing files.
API import works by re-creating the thumbnails when it transfers over files. Only the original file is downloaded and it’s inserted more or less like any other file. Generally simply copying over thumbnails shouldn’t work, for a couple reasons: first, Omeka flags in the database whether thumbnails were created for a file, and if that flag is not set, it won’t look for stored thumbnails at all, and second, the randomized filename used for the original and thumbnails when stored will be different from one server to the other.
I’m not aware of anything specifically related to API Import that would make imported files not generate thumbnails. It’s possible that there was a problem with thumbnail generation on your site at the time you did the import, which has now been fixed. Or, possibly that the imported files that didn’t get thumbnails differ in some way from the ones you’re successfully doing with new uploads (for example, if the imported files were all PDFs and you’re testing JPGs now).
As for the “copying over” working, if you weren’t accounting for the issues I mentioned, I can’t think of a way where simply copying would work if you indeed did do an API import. For a more manual process like copying the database directly from another install, it would make some sense, but not an import.
I was actually switching that flag in database so it would look for the thumbnails. I’m going to abandon this approach in favor of a fresh import. Thanks for the responses.