Use initial schema during bulk load.

[martinmr]

Use the initial schema during bulk load, otherwise data might not be
loaded correctly for those predicates.

In particular, I observed that when loading the 21million dataset with
dgraph.type triples added to it, queries would not respond and would
eventually timeout. I figured this was because the index for that
predicate was not built during the bulkload. I reloaded the dataset with
my changes included and queries respond immediately.

Fixes #3329

---

This change is 

[codexnull]

Reviewed 2 of 2 files at r1.
Reviewable status: complete! all files reviewed, all discussions resolved (waiting on @manishrjain)

[manishrjain]

Reviewable status: all files reviewed, 1 unresolved discussion (waiting on @martinmr)

---

schema/schema.go, line 423 at r1 (raw file):

// them later than miss some of them. An example of such situation is during bulk
// loading.
func CompleteInitialSchema() []*pb.SchemaUpdate {

So, does it mean that bulk loader would always spit out the schemas for acls, even when the user only wants the open source version?

[martinmr]

Reviewable status: all files reviewed, 1 unresolved discussion (waiting on @manishrjain)

---

schema/schema.go, line 423 at r1 (raw file):

Previously, manishrjain (Manish R Jain) wrote…

So, does it mean that bulk loader would always spit out the schemas for acls, even when the user only wants the open source version?

Initial schema will consider the worker options when deciding if a predicate should be created at start up. Right now there are two type of optional predicates (_predicate_ and ACL predicates) but there might be more in the future so I thought it would be safer to add all of them during the bulk process rather than risking missing some of them and not creating the proper indices during bulk load.

In any case, the ACL predicates are considered reserved even if ACL is off, so users should not be able to create another predicate with the same name.

[manishrjain]

Reviewed 1 of 2 files at r1.
Reviewable status: all files reviewed, 1 unresolved discussion (waiting on @manishrjain)