Unable to create shared cache for several instances of the same http-service

Hello.

We are running a several instances of a service with shared redis-cache and we found some concurrency problems.

If I understood the logic correctly, it is like this:

1. Check the cache.
2. If there is a cached value, then return it.
3. If not, then create a writer and run handler.
4. When handler is done, write response to cache.
5. The most unobvious part: if there is a cached value (so it must have appeared during step 3), append the current response data to it, instead of overwriting it.
   https://github.com/gin-contrib/cache/blob/6b4ffed64d167edff23695c7bd60af88a6792120/cache.go#L84-L86
6. ???
7. PROFIT

So, when we send two equal queries and our balancer sends them to different instances, we will get the problem, mentioned in this closed issue: https://github.com/gin-contrib/cache/issues/15 Because two instances will both run handlers and the first one to finish will add the value to the cache and the second one will append it again.

What is the logic behind step 5? Maybe, we are using this lib incorrectly?

Thank you.

Read cache.go around lines 84-86 and compare the behavior with the closed issue #15. Reproduce two equal requests against separate service instances using the shared Redis cache, then establish the expected result and document or test the behavior so completion is verifiable.