files_by_digest API call could be taking too long

[jcoyne]

Here is an example of timeouts in DSA when calling that endpoint: https://app.honeybadger.io/projects/50568/faults/110740789

This appears to be the /v1/purls/{druid} route. https://github.com/sul-dlss/purl-fetcher/blob/main/app/controllers/v1/purls_controller.rb#L17

Which calls FilesByMd5Service.call(purl:)

[jcoyne]

For some of these (druid:tm782sf2963) I'm seeing 15s, which is long, but I wouldn't think so long as to cause the timeout:

puts Benchmark.measure { FilesByMd5Service.call(purl:) }
  2.004958   0.742488   2.747446 ( 14.504378)

druid:sq409qv3664 was a bit longer, but still not awful:

puts Benchmark.measure { FilesByMd5Service.call(purl:) }
  2.738115   1.334335   4.072450 ( 27.016673)

There does appear to be some caching, as the first call to this function seems to take about 10s longer than subsequent calls.

Sometimes the purl that was failing is now super fast, which makes me thing it's a network issue:

3.3.0 :020 > purl = Purl.find_by(druid: 'druid:td587yz4669')
 =>
#<Purl:0x00007fbf955ba8e0
...
3.3.0 :021 > puts Benchmark.measure { FilesByMd5Service.call(purl:) }
  0.018891   0.000878   0.019769 (  0.129804)