Looking into similiar issues on our end right now. If you find something, let me know please!

this looks quite crazy, I will do some more tests tomorrow to see if it is really related to the cleanup method in node-cache.

@simllll , Did you set TTL as 1min?

TTL was different, but checkperiod was set to 60. I tried now setting it to 0 (to disable the check), but turned out it didn't make any difference. Therefore I assume this is not the real issue here.

Can you try setting checkperiod to 0 on your side to see if it changes something?

Simon

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

One possible optimization would be to dereference all the expired keys in one statement vs while looping over them. By keeping a reference to all the to-be-evicted-soon objects in one place until we're about to be done with the check and dereferencing them en masse would mean that we trigger one GC vs a dice roll on each dereference on each iteration of the loop.

REF: https://stackoverflow.com/a/52745907/968201

The API of check could change. Right now check actually does the delete. Instead it could return a boolean of if the key should be evicted. Then we could collect an array of keys to evict in checkData. After looping we could do one call to del with all the keys to be evicted. We might also want to change the the del code to keep one final reference from key -> oldVal. Which will dereference when the function returns triggering 1 GC (if the engine wants to run it)