-
Notifications
You must be signed in to change notification settings - Fork 4.5k
Conversation
8242a02
to
558a4e5
Compare
558a4e5
to
353ab28
Compare
353ab28
to
9265214
Compare
Codecov Report
@@ Coverage Diff @@
## master #21322 +/- ##
=========================================
- Coverage 81.5% 81.5% -0.1%
=========================================
Files 500 501 +1
Lines 140511 140590 +79
=========================================
+ Hits 114655 114664 +9
- Misses 25856 25926 +70 |
let mut dropped_banks_time = Measure::start("drop_banks"); | ||
drop(banks); | ||
dropped_banks_time.stop(); | ||
if dropped_banks_time.as_ms() > 10 { |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Maybe worth logging a datapoint for each individual bank for each slot in case it's something bank specific?
Also not a big deal, but I think it's possible some other thread has an outstanding reference to some Arc
root bank even after it was removed in the replay path (something like, RPC or the vote listener) but this at least unblocks replay.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I've been thinking along the same lines as Carl, but unblocking replay is still good.
Thanks for pulling the metrics
(cherry picked from commit 0bda0c3) # Conflicts: # core/src/lib.rs # core/src/replay_stage.rs # core/src/tvu.rs # core/src/vote_simulator.rs
(cherry picked from commit 0bda0c3)
(cherry picked from commit 0bda0c3)
(cherry picked from commit 0bda0c3)
(cherry picked from commit 0bda0c3)
This reverts commit d379f09.
This reverts commit b030d4b.
Problem
Dropping banks can be expensive and it's bad to stall the replay stage loop.
Summary of Changes
Drop banks in another thread.
Fixes #