You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
And does the eviction handler then block data or threads for such a long time that the end-user observes a very long response time?
If that is the case, would it not be more handy to give the user some control over this behavior? Either beforehand by giving the run_pending_tasks() function an input u64 parameter of the maximum amount of items which can processed, and/or afterwards supply a boolean or u64 as return value containing the numbers of items which still needs to be processed...?
So, without queued delivery mode, cache writes could be blocked in the above case. As you see in the migration guide, run_pending_tasks does many types of task, so we could provide a struct with some fields to specify the maximum amounts of items to process per task type. As for the return type, it is expensive (e.g. O(n)) to count the numbers of items, which still needs to be processed. So probably we will return a stuct having the counters for the numbers of items, which were processed, per task type.