At time of fork, it was part of the semi-frequent maintenance process to delete every key in Redis. This is because some uses of flask-caching did not have cache timeouts set. Or, for places like the logged-in counter that store data directly, were not pruning old entries.
Redis will soon be persisted across restarts, and this issue is timely and likely needs resolved within a couple weeks before performance issues accumulate (diagnosable by a sampling profiler spending >20% time in json.loads).
At time of fork, it was part of the semi-frequent maintenance process to delete every key in Redis. This is because some uses of flask-caching did not have cache timeouts set. Or, for places like the logged-in counter that store data directly, were not pruning old entries.
Redis will soon be persisted across restarts, and this issue is timely and likely needs resolved within a couple weeks before performance issues accumulate (diagnosable by a sampling profiler spending >20% time in
json.loads
).