Details
-
Improvement
-
Resolution: Fixed
-
Minor
-
5.5.0
Description
From a quick review of memcached logging in Vulcan it looks like a new 'slow runtime' warning was added for tasks, which is really handy.
Unfortunately it seems quite aggressive:
2018-03-14T13:59:29.776326Z WARNING (travel-sample) Slow runtime for 'Checkpoint Remover on vb 707' on thread nonIO_worker_2: 15 ms
|
2018-03-14T13:59:29.729062Z WARNING (travel-sample) Slow runtime for 'Checkpoint Remover on vb 1014' on thread nonIO_worker_0: 15 ms
|
I understand that these may usually be expected to run very quickly (this is only on a 2-core VM), but my concern is that these 'slow runtimes' may cause logs to wrap very quickly, particularly in undersized environments.
Could we make these thresholds more conservative while also ensuring that they trigger in the correct circumstances to be useful?
For example we did something similar with the cmd_compact command in MB-16003.