diff --git a/contrib/scripts/test-bulk-schema.sh b/contrib/scripts/test-bulk-schema.sh
new file mode 100755
index 00000000000..c9fa8d1d215
--- /dev/null
+++ b/contrib/scripts/test-bulk-schema.sh
@@ -0,0 +1,227 @@
+#!/bin/bash
+# verify fix of https://github.com/dgraph-io/dgraph/issues/2616
+
+readonly ME=${0##*/}
+readonly SRCDIR=$(readlink -f ${BASH_SOURCE[0]%/*})
+
+declare -ri PORT_OFFSET=$((RANDOM % 1000))
+declare -ri ZERO_PORT=$((5080+PORT_OFFSET))
+declare -ri ALPHA_PORT=$((7080+PORT_OFFSET)) HTTP_PORT=$((8080+PORT_OFFSET))
+
+INFO() { echo "$ME: $@"; }
+ERROR() { echo >&2 "$ME: $@"; }
+FATAL() { ERROR "$@"; exit 1; }
+
+set -e
+
+INFO "running bulk load schema test"
+
+WORKDIR=$(mktemp --tmpdir -d $ME.tmp-XXXXXX)
+INFO "using workdir $WORKDIR"
+cd $WORKDIR
+
+function StartZero
+{
+  INFO "starting zero server on port $ZERO_PORT"
+  dgraph zero -o $PORT_OFFSET --my=localhost:$ZERO_PORT \
+    >zero.log 2>&1 </dev/null &
+  ZERO_PID=$!
+  sleep 1
+  $SRCDIR/../wait-for-it.sh -q -t 30 localhost:$ZERO_PORT \
+    || FATAL "failed to start zero"
+}
+
+function BulkLoadSampleData
+{
+  INFO "bulk loading sample data"
+  cat >1million.schema <<EOF
+director.film: uid @reverse .
+genre: uid @reverse .
+initial_release_date: dateTime @index(year) .
+name: string @index(term) @lang .
+EOF
+  mkfifo 1million.rdf.gz
+  curl -LsS 'https://github.com/dgraph-io/tutorial/blob/master/resources/1million.rdf.gz?raw=true' >> 1million.rdf.gz &
+  dgraph bulk -z localhost:$ZERO_PORT -s 1million.schema -r 1million.rdf.gz \
+     >bulk.log 2>&1 </dev/null
+}
+
+function StartAlpha
+{
+  INFO "starting alpha server on port $ALPHA_PORT"
+  dgraph alpha -o $PORT_OFFSET --my=localhost:$ALPHA_PORT --zero=localhost:$ZERO_PORT --lru_mb=2048 \
+      >alpha.log 2>&1 </dev/null &
+  ALPHA_PID=$!
+  sleep 1
+  $SRCDIR/../wait-for-it.sh -q -t 30 localhost:$ALPHA_PORT \
+    || FATAL "failed to start alpha"
+}
+
+function UpdateDatabase
+{
+  INFO "adding predicate with default type to schema"
+  curl localhost:$HTTP_PORT/alter -X POST -d$'
+predicate_with_no_uid_count:string  .
+predicate_with_default_type:default  .
+predicate_with_index_no_uid_count:string @index(exact) .
+' &>/dev/null
+
+  curl localhost:$HTTP_PORT/mutate -X POST -H 'X-Dgraph-CommitNow: true' -d $'
+{
+  set {
+    _:company1 <predicate_with_default_type> "CompanyABC" .
+  }
+}
+' &>/dev/null
+}
+
+function QuerySchema
+{
+  INFO "running schema query"
+  local out_file=${1:?no out file}
+  curl -sS localhost:$HTTP_PORT/query -XPOST -d'schema {}' | python -c "import json,sys; d=json.load(sys.stdin); json.dump(d['data'],sys.stdout,sort_keys=True,indent=2,separators=(',',': '))" > $out_file
+  echo >> $out_file
+  #INFO "schema is: " && cat $out_file
+}
+
+function DoExport
+{
+  INFO "running export"
+  curl localhost:$HTTP_PORT/admin/export &>/dev/null
+  sleep 1
+}
+
+function BulkLoadExportedData
+{
+  INFO "bulk loading exported data"
+  dgraph bulk -z localhost:$ZERO_PORT \
+              -s ../dir1/export/*/g01.schema.gz \
+              -r ../dir1/export/*/g01.rdf.gz \
+     >bulk.log 2>&1 </dev/null
+  mv out/0/p .
+}
+
+function BulkLoadFixtureData
+{
+  INFO "bulk loading fixture data"
+
+  # schema test cases:
+  #
+  # 1. predicate with non-default type (name)
+  # 2. predicate with default type (genre)
+  # 3. predicate not used in rdf (language)
+  cat >fixture.schema <<EOF
+name:string @index(term) .
+genre:default .
+language:string .
+EOF
+
+  # rdf test cases:
+  #
+  # 4. predicate not in schema (revenue)
+  cat >fixture.rdf <<EOF
+_:et <name> "E.T. the Extra-Terrestrial" .
+_:et <genre> "Science Fiction" .
+_:et <revenue> "792.9" .
+EOF
+
+  dgraph bulk -z localhost:$ZERO_PORT -s fixture.schema -r fixture.rdf \
+     >bulk.log 2>&1 </dev/null
+  mv out/0/p .
+}
+
+function StopServers
+{
+  INFO "killing zero server at pid $ZERO_PID"
+  INFO "killing alpha server at pid $ALPHA_PID"
+  kill $ZERO_PID $ALPHA_PID
+  sleep 1
+}
+
+function Cleanup
+{
+  INFO "removing $WORKDIR"
+  rm -rf $WORKDIR
+}
+
+mkdir dir1
+pushd dir1 >/dev/null
+
+StartZero
+BulkLoadSampleData
+StartAlpha
+UpdateDatabase
+QuerySchema "schema.out"
+DoExport
+StopServers
+
+popd >/dev/null
+mkdir dir2
+pushd dir2 >/dev/null
+
+StartZero
+BulkLoadExportedData
+StartAlpha
+QuerySchema "schema.out"
+StopServers
+
+popd >/dev/null
+
+INFO "verifing schema is same before export and after bulk import"
+diff dir1/schema.out dir2/schema.out || FATAL "schema incorrect"
+INFO "schema is correct"
+
+mkdir dir3
+pushd dir3 >/dev/null
+
+StartZero
+BulkLoadFixtureData
+StartAlpha
+QuerySchema "schema.out"
+StopServers
+
+popd >/dev/null
+
+# final schema should include *all* predicates regardless of whether they were
+# introduced by the schema or rdf file, used or not used, or of default type
+# or non-default type
+INFO "verifying schema contains all predicates"
+diff - dir3/schema.out <<EOF || FATAL "schema incorrect"
+{
+  "schema": [
+    {
+      "list": true,
+      "predicate": "_predicate_",
+      "type": "string"
+    },
+    {
+      "predicate": "genre",
+      "type": "default"
+    },
+    {
+      "predicate": "language",
+      "type": "string"
+    },
+    {
+      "index": true,
+      "predicate": "name",
+      "tokenizer": [
+        "term"
+      ],
+      "type": "string"
+    },
+    {
+      "predicate": "revenue",
+      "type": "default"
+    }
+  ]
+}
+EOF
+
+INFO "schema is correct"
+
+Cleanup
+
+exit 0
+
+# eof
diff --git a/dgraph/cmd/bulk/run.go b/dgraph/cmd/bulk/run.go
index 09fc9a82a8b..9bd4fe9048a 100644
--- a/dgraph/cmd/bulk/run.go
+++ b/dgraph/cmd/bulk/run.go
@@ -18,7 +18,6 @@ package bulk
 
 import (
 	"encoding/json"
-	"flag"
 	"fmt"
 	"log"
 	"net/http"
@@ -113,7 +112,6 @@ func run() {
 		os.Exit(0)
 	}
 	if opt.RDFDir == "" || opt.SchemaFile == "" {
-		flag.Usage()
 		fmt.Fprint(os.Stderr, "RDF and schema file(s) must be specified.\n")
 		os.Exit(1)
 	}
diff --git a/dgraph/cmd/bulk/schema.go b/dgraph/cmd/bulk/schema.go
index e29d6686ca0..af220129398 100644
--- a/dgraph/cmd/bulk/schema.go
+++ b/dgraph/cmd/bulk/schema.go
@@ -120,7 +120,21 @@ func (s *schemaStore) getPredicates(db *badger.DB) []string {
 func (s *schemaStore) write(db *badger.DB) {
 	// Write schema always at timestamp 1, s.state.writeTs may not be equal to 1
 	// if bulk loader was restarted or other similar scenarios.
-	preds := s.getPredicates(db)
+
+	// Get predicates from the schema store so that the db includes all
+	// predicates from the schema file.
+	preds := make([]string, 0, len(s.m))
+	for pred := range s.m {
+		preds = append(preds, pred)
+	}
+
+	// Add predicates from the db so that final schema includes predicates
+	// used in the rdf file but not included in the schema file.
+	for _, pred := range s.getPredicates(db) {
+		if _, ok := s.m[pred]; ! ok {
+			preds = append(preds, pred)
+		}
+	}
 
 	txn := db.NewTransactionAt(math.MaxUint64, true)
 	defer txn.Discard()
diff --git a/schema/schema.go b/schema/schema.go
index 21f98c0dc0b..f10cc01536d 100644
--- a/schema/schema.go
+++ b/schema/schema.go
@@ -282,7 +282,7 @@ func LoadFromDb() error {
 		var s pb.SchemaUpdate
 		err := item.Value(func(val []byte) error {
 			if len(val) == 0 {
-				return nil
+				s = pb.SchemaUpdate{Predicate: attr, ValueType: pb.Posting_DEFAULT}
 			}
 			x.Checkf(s.Unmarshal(val), "Error while loading schema from db")
 			State().Set(attr, s)
diff --git a/test.sh b/test.sh
index 723f1fc4116..2efbccebcc7 100755
--- a/test.sh
+++ b/test.sh
@@ -1,6 +1,9 @@
 #!/bin/bash
 
-source contrib/scripts/functions.sh
+# run from directory containing this script
+cd ${BASH_SOURCE[0]%/*}
+
+source ./contrib/scripts/functions.sh
 function run {
   go test -short=true $@ |\
 		GREP_COLORS='mt=01;32' egrep --line-buffered --color=always '^ok\ .*|$' |\
@@ -31,6 +34,9 @@ echo
 echo "Running tests. Ignoring vendor folder."
 runAll || exit $?
 
+# Run non-go tests.
+./contrib/scripts/test-bulk-schema.sh
+
 echo
 echo "Running load-test.sh"
 ./contrib/scripts/load-test.sh