Description
What is the problem?
After Oct. 8, 2023, CDN cache misses increased by 7x
, from 0.75m
to 5.2m
This seems to be coming from the app origin, specifically F_fly_2022_03_13
If we group by url
, we notice that /feed
, /
& /feed/
requests are the ones with the highest increase in cache misses:

What is the Fastly response if we enable debug headers?

This looks like a clean cache miss, and we do not have shielding enabled. Fixed it.
Note
While at it, I did the same for the feeds origin.
Does this fix it?
This looks a lot better now:

👉 Let's give this a few hours before we check the hits vs misses again.
Questions
Do the cache headers look correct?
Looking at a specific URL endpoint served by the origin, we see the following:

This looks correct to me. surrogate-control
is set and should have precedence over cache-control
. Reference: Understand how cache control headers work.
If the surrogate-control
headers are respected, why does this content have so many misses? cc @jamesarosen
Could this config change be related to the increased cache misses?
We had a two VCL changes earlier that day:
The second one looks suspicious, but on closer inspection we are simply setting surrogate-control
to AWS S3 headers which are prefixed with x-amz-meta-
:
Could the above change be related to the increased cache misses @jamesarosen?
To be continued...
Activity