Cloudflare's Weird Age Header.
I noticed something weird today:
% curl -I https://blog.benjojo.co.uk/rss.xml HTTP/2 200 age: 99788 cache-control: public, max-age=1800 cf-cache-status: HIT date: Fri, 18 Feb 2022 17:41:40 GMT server: cloudflare ...
Do you see the oddity?
Age is greater than
max-age. According to RFC2616 that means that this resource is already expired. Let’s look at the “Summary of age calculation algorithm”.
apparent_age = max(0, response_time - date_value) corrected_received_age = max(apparent_age, age_value) response_delay = response_time - request_time corrected_initial_age = corrected_received_age + response_delay resident_time = now - response_time current_age = corrected_initial_age + resident_time
Let’s fill in some variables and evaluate…
age_value = 101626 date_value = "2022-02-18 17:41:40" now = "2022-02-18 17:41:41" request_time = "2022-02-18 17:41:39" response_time = "2022-02-18 17:41:41" apparent_age = max(0, "2022-02-18 17:41:41" - "2022-02-18 17:41:40") # 1 corrected_received_age = max(1, 101626) # 101626 response_delay = "2022-02-18 17:41:41" - "2022-02-18 17:41:39" # 2 corrected_initial_age = 101626 + 2 # 101628 resident_time = "2022-02-18 17:41:41" - "2022-02-18 17:41:41" # 0 current_age = 101628 + 0 # 101628
I’ll save you the math for calculating freshness but it boils down to
current_age <= max_age AKA
101628 <= 1800 == false # STALE.
This means that Cloudflare is serving resources that are already stale! Now this isn’t much of a problem, the customer may have configured Cloudflare to cache for an extended period of time, ignoring the original headers. However in that case Cloudflare should probably “spoof” the headers to reflect reality. In this case this the best option is likely updating the
Cache-Control header to suggest that the response still has a little life left in it. Or just set
max-age=0 to indicate that it needs to be revalidated, I think this option would preserve the current behaviour while being less confusing.
One explanation is that the origin is down. The
Age header is over 28h old, so it does seem unlikely that this was an intentional cache period. In this case they may be using something like a
s-stale-if-error extension to indicate that the cache may use expired entries if the origin is down.
Another possible explanation is that Cloudflare revalidated the cache entry, however in this case the
Age header should be reset to the time of validation.
I don’t think this is a major problem, the only downside I can think of is that it will lead to clients constantly contacting Cloudflare to revalidate, which may not have been the original intention. However it certainly did pique my curiosity.
Feed - Newsletter - Private Feedback