We have an issue in one of the projects. tnt_thread_cpu metrics, which has only one counter and pid (or else) for a label generated enormous amounts of timelines.
this may cause memory issues in a monitoring system.
Looks like labels are left even after cpu thread has finished its work. But there is no way to remove the clutter.
We have an issue in one of the projects.
tnt_thread_cpu
metrics, which has only one counter andpid
(or else) for a label generated enormous amounts of timelines.this may cause memory issues in a monitoring system.
Looks like labels are left even after cpu thread has finished its work. But there is no way to remove the clutter.