[Enhancement] MetricEntity busy wait during memtable flush

[platoneko]

Search before asking

• I had searched in the issues and found no similar issues.

Description

Currently, MetricRegistry and MetricEntity use SpinLock to solve concurrency problems. However when MetricRegistry call trigger_all_hooks, these triggered hook functions may acquire other mutex, and these mutex may be held during some long time work.


i.e. LoadChannelMgr::_lock may be held during MemTable flush.

Solution

No response

Are you willing to submit PR?

• Yes I am willing to submit a PR!

Code of Conduct

• I agree to follow this project's Code of Conduct

[platoneko]

Or can we avoid acquiring lock in metrics hook functions?