For a caching library to do its part to mitigate distributed outages, it's a best practice to cache exception responses with a distinct (usually shorter) expiry. Without this, an outage in a backing service can cause errors in cached functions, which result in an increased number of calls going to the backing service that is struggling, since calls that would normally leave a cached result are now exiting without writing to the cache. Caching exceptions will keep load off of the service the backing service that is struggling.
Inspired by recommendations in https://aws.amazon.com/builders-library/caching-challenges-and-strategies.
For a caching library to do its part to mitigate distributed outages, it's a best practice to cache exception responses with a distinct (usually shorter) expiry. Without this, an outage in a backing service can cause errors in cached functions, which result in an increased number of calls going to the backing service that is struggling, since calls that would normally leave a cached result are now exiting without writing to the cache. Caching exceptions will keep load off of the service the backing service that is struggling.
Inspired by recommendations in https://aws.amazon.com/builders-library/caching-challenges-and-strategies.