images not displaying on images (iiif?)

davidschober commented 5 years ago

Severity

Is the production site running?

[x] yes
[ ] no

Are staff blocked from performing their work?

[x] yes
[ ] no

Descriptive summary

Some images and / or thumbnails are not displaying on image.northwestern.edu. It's possible the images that are displaying are already cached by RIIIF.

Images that do not display error when hitting the iiif server directly https://images.northwestern.edu/image-service/inu-dil-cea91f6b-3217-40ea-a8ff-d8b8d3e2e6ad/full/full/0/default.jpg

Steps to reproduce the behavior

Go to an image on images that doesn't display http://images.northwestern.edu/multiresimages/inu:dil-4678779e-03e3-4093-a9f3-b60df4b93f32
see that it doesn't display
verify it by looking at the iiif endpoint of the image http://images.northwestern.edu/image-service/inu-dil-4678779e-03e3-4093-a9f3-b60df4b93f32/full/,1600/0/default.jpg
Compare to one that workshttp://images.northwestern.edu/multiresimages/inu:dil-67a05fd7-b19e-4bfc-93e1-7738353f0b71 http://images.northwestern.edu/image-service/inu-dil-67a05fd7-b19e-4bfc-93e1-7738353f0b71/full/,1600/0/default.jpg

davidschober commented 5 years ago

A few different things happened here.

There were some errant jp2s which when the deep cataloged was searched caused the server to crash. We are running a script to convert those to TIFs.
A bunch of images were cached with bad keys. We were able to clear out the bad keys

Slack discussion

mbklein [10:54 AM] I’ll check

Nope. Never mind. I don’t need to. Ugh. The key is based on the MD5 of the URL, which is going to be impossible to figure out based on the URL prefix. Maybe I can do that one jpg though.

mbklein [11:08 AM] Almost there.

mbklein [11:19 AM] Success http://images.northwestern.edu/image-service/inu-dil-5f4c79ec-9f65-4e41-88aa-b4f1e128829b/full/,120/0/default.jpg http://images.northwestern.edu/image-service/inu-dil-5f4c79ec-9f65-4e41-88aa-b4f1e128829b/full/,120/0/default.jpg

david.schober [11:19 AM] whoot! So we know it's the cache

mbklein [11:21 AM] Can you find me another bad one? I want to check something and I’ve already blown away that one. :slightly_smiling_face:

david.schober [11:21 AM] Sure! There's a lot of 'em it's like fishing in the north woods http://images.northwestern.edu/image-service/inu-dil-c9c1ab35-8cfe-433e-91f4-a4749367216b/full/,120/0/default.jpg

mbklein [11:23 AM] Aha! Progress!

david.schober [11:23 AM] Frankly I think its everything that I preprimed (first page of all public collections) which fits nicely into the Karen Theory of non-randomized Image Entropy (edited)

mbklein [11:28 AM] I found the commonality. It’s easy once you know an item’s cache key.

david.schober [11:29 AM] come on man, you're killing me!

mbklein [11:29 AM] Rails.cache.fetch(key) is an empty string for the busted items. I’ve cleared out just those 224 items.

davidschober commented 5 years ago

FOR THE RECORD:

def riiif_cache_key_for(url)
  opts = {}
  (id, opts[:region], opts[:size], opts[:rotation], opts[:quality], opts[:format]) = URI.parse(url).path.split(%r{[/.]})[-6..-1]
  decoded = Riiif::Image.new(id).send(:decode_options!, opts)
  Riiif::Image.cache_key(id, decoded).gsub(/:/,'%3A')
end

nulib / images