Cancel node context after StopAll #8289
Conversation
rkapka
changed the title
slasher/node/node.go
Outdated
| @@ -150,11 +150,11 @@ func (s *SlasherNode) Close() { | |||
| defer s.lock.Unlock() | |||
|
|
|||
| log.Info("Stopping hash slinging slasher") | |||
| s.cancel() | |||
There was a problem hiding this comment.
The same bug-prone code as in the beacon node - cancelling the parent context before StopAll is called.
Codecov Report
@@ Coverage Diff @@
## develop #8289 +/- ##
==========================================
Coverage ? 57.12%
==========================================
Files ? 449
Lines ? 31524
Branches ? 0
==========================================
Hits ? 18007
Misses ? 10719
Partials ? 2798 |
|
Conflict and tests failing |
rkapka
dismissed stale reviews from prestonvanloon and rauljordan
via
3dc942b
It's the duplicate wallet import test that is failing. It's causing problems in all of our PRs. |
There was a problem hiding this comment.
How about validator?
https://github.com/prysmaticlabs/prysm/blob/master/validator/node/node.go#L151
I had not implemented this for validator because the struct did not contain |
What type of PR is this?
Other
What does this PR do? Why is it needed?
#7625 removed context cancelling from the beacon node because it caused a panic - cancelling the parent context before
StopAll, during which services are being shut down, caused child contexts inside these services to be cancelled before services had a chance to clean themselves up.Last week me and @prestonvanloon discussed some related issue and it occurred to us that it's safe to cancel the parent context after all child contexts have already done their job, i.e. after
StopAll. This PR adopts this solution.Which issues(s) does this PR fix?
N/A
Other notes for review
So simple yet so clever.