Use MaxAssigned for ReadOnly TS

[srfrog]

This change will use the stored MaxAssigned from ProcessDelta() as read-only timestamp. This will save an extra network call to Zero.

---

This change is 

[manishrjain]

We need to add another bool to mention "best effort". We shouldn't change the meaning of existing read-only txns. This would require a change at the client side and in Request proto as well.

Reviewed 2 of 2 files at r1.
Reviewable status: all files reviewed, 2 unresolved discussions (waiting on @srfrog)

---

worker/mutation.go, line 339 at r1 (raw file):

	if num.Val == 0 && num.ReadOnly {
		if readTs := posting.Oracle().MaxAssigned(); readTs > 0 {
			glog.V(3).Infof("... Timestamp served from memory: %v", readTs)

Drop the V(3) log.

---

worker/mutation.go, line 340 at r1 (raw file):

		if readTs := posting.Oracle().MaxAssigned(); readTs > 0 {
			glog.V(3).Infof("... Timestamp served from memory: %v", readTs)
			return &pb.AssignedIds{ReadOnly: readTs + 1}, nil

This should not be a one above MaxAssigned. It should be at MaxAssigned.

[srfrog]

Reviewable status: all files reviewed, 2 unresolved discussions (waiting on @manishrjain)

---

worker/mutation.go, line 339 at r1 (raw file):

Previously, manishrjain (Manish R Jain) wrote…

Drop the V(3) log.

Done.

---

worker/mutation.go, line 340 at r1 (raw file):

Previously, manishrjain (Manish R Jain) wrote…

This should not be a one above MaxAssigned. It should be at MaxAssigned.

Done.

[manishrjain]

dgraph/edgraph/server.go

Line 572 in fde781a

if req.StartTs == 0 {

^ This is right place to make a change. If req has best effort (and must also have read only, or we return an error), then here we can get a timestamp from Oracle().MaxAssigned().

Reviewable status: 1 of 5 files reviewed, 5 unresolved discussions (waiting on @manishrjain and @srfrog)

---

edgraph/server.go, line 233 at r2 (raw file):

			if r.readOnly {
				num.ReadOnly = true
				bestEffort = r.bestEffort

we shouldn't be sending a best effort request here.

---

edgraph/server.go, line 268 at r2 (raw file):

type tsReq struct {
	readOnly, bestEffort bool

Not required.

---

worker/mutation.go, line 337 at r2 (raw file):

func Timestamps(ctx context.Context, num *pb.Num, bestEffort bool) (*pb.AssignedIds, error) {
	if num.Val == 0 && num.ReadOnly && bestEffort {

shouldn't be required to do it here.

[manishrjain]

This change as it stands isn't required. Probably better to start from scratch.

Reviewable status: 1 of 5 files reviewed, 5 unresolved discussions (waiting on @manishrjain and @srfrog)

[manishrjain]

Reviewed 2 of 4 files at r3.
Reviewable status: all files reviewed, 7 unresolved discussions (waiting on @manishrjain and @srfrog)

---

edgraph/server.go, line 574 at r3 (raw file):

	switch {
	case req.StartTs == 0:

Not worth a switch case, if it only has one case.

---

edgraph/server.go, line 582 at r3 (raw file):

			}
			if readTs := posting.Oracle().MaxAssigned(); readTs > 0 {
				glog.Infof("Timestamp served from memory: %v", readTs)

Remove the infof.

---

edgraph/server.go, line 587 at r3 (raw file):

			}
			// Our best effort failed, fall back to normal mode.
		}

This should be else here.

[srfrog]

Reviewable status: 1 of 2 files reviewed, 7 unresolved discussions (waiting on @manishrjain)

---

edgraph/server.go, line 233 at r2 (raw file):

Previously, manishrjain (Manish R Jain) wrote…

we shouldn't be sending a best effort request here.

Done.

---

edgraph/server.go, line 268 at r2 (raw file):

Previously, manishrjain (Manish R Jain) wrote…

Not required.

Done.

---

edgraph/server.go, line 574 at r3 (raw file):

Previously, manishrjain (Manish R Jain) wrote…

Not worth a switch case, if it only has one case.

Done.

---

edgraph/server.go, line 582 at r3 (raw file):

Previously, manishrjain (Manish R Jain) wrote…

Remove the infof.

Done.

---

edgraph/server.go, line 587 at r3 (raw file):

Previously, manishrjain (Manish R Jain) wrote…

This should be else here.

Done.

---

worker/mutation.go, line 337 at r2 (raw file):

Previously, manishrjain (Manish R Jain) wrote…

shouldn't be required to do it here.

Done.

[manishrjain]

Reviewed 1 of 1 files at r4.
Reviewable status: all files reviewed, 1 unresolved discussion (waiting on @manishrjain)