CommitteeDutyGuard: share between committees

[iurii-ssv]

Hopefully closes #1745

I've been looking into CommitteeDutyGuard logic and got a question (see below).

Decided to take a swing at related issue while I'm at it :)

[iurii-ssv]

Just a note, for operators who serve many validators CommitteeDutyGuard.mtx could be somewhat contentious,

it's unlikely to be an issue, but if it is - we can optimize this (later on) by having a bunch of mutexs (each serving a subset/bucket of validators).

[moshe-blox]

@iurii-ssv operators can serve up to 500 validators at the moment, which comes off at an average of 15 per slot (15 at a time), so it should be insignificant given each call should finish within nanoseconds

of course, if this surprisingly pops up on pprof, then we may consider a different approach

[y0sher]

Great!

[iurii-ssv]

Corresponding diff:

diff --git a/protocol/v2/ssv/validator/committee.go@NewCommittee b/ssv/committee.go@NewCommittee
index a347ee50e92f7334..2222222
--- a/protocol/v2/ssv/validator/committee.go@NewCommittee
+++ b/ssv/committee.go@NewCommittee
@@ -1,25 +1,13 @@
 func NewCommittee(
-	ctx context.Context,
-	cancel context.CancelFunc,
-	beaconNetwork BeaconNetwork,
-	committeeMember *CommitteeMember,
-	createRunnerFn CommitteeRunnerFunc,
-	shares map[phase0.ValidatorIndex]*Share,
-	dutyGuard *CommitteeDutyGuard,
+	committeeMember CommitteeMember,
+	share map[spec.ValidatorIndex]*Share,
+	createRunnerFn CreateRunnerFn,
 ) *Committee {
-	if shares == nil {
-		shares = make(map[phase0.ValidatorIndex]*Share)
-	}
-	return &Committee{
-		logger:			logger,
-		BeaconNetwork:		beaconNetwork,
-		ctx:			ctx,
-		cancel:			cancel,
-		Queues:			make(map[phase0.Slot]queueContainer),
-		Runners:		make(map[phase0.Slot]*runner.CommitteeRunner),
-		Shares:			shares,
+	c := &Committee{
+		Runners:		make(map[spec.Slot]*CommitteeRunner),
 		CommitteeMember:	committeeMember,
 		CreateRunnerFn:		createRunnerFn,
-		dutyGuard:		dutyGuard,
+		Share:			share,
 	}
+	return c
 }
\ No newline at end of file

[iurii-ssv]

This PR seems to be working fine on stage.