diff --git a/src/backend/columnar/columnar_tableam.c b/src/backend/columnar/columnar_tableam.c
index 944032b84..3ed5d8187 100644
--- a/src/backend/columnar/columnar_tableam.c
+++ b/src/backend/columnar/columnar_tableam.c
@@ -343,21 +343,33 @@ ErrorIfInvalidRowNumber(uint64 rowNumber)
 static Size
 columnar_parallelscan_estimate(Relation rel)
 {
-	elog(ERROR, "columnar_parallelscan_estimate not implemented");
+	return sizeof(ParallelBlockTableScanDescData);
 }
 
 
 static Size
 columnar_parallelscan_initialize(Relation rel, ParallelTableScanDesc pscan)
 {
-	elog(ERROR, "columnar_parallelscan_initialize not implemented");
+	ParallelBlockTableScanDesc bpscan = (ParallelBlockTableScanDesc) pscan;
+
+	bpscan->base.phs_relid = RelationGetRelid(rel);
+	bpscan->phs_nblocks = RelationGetNumberOfBlocks(rel);
+	bpscan->base.phs_syncscan = synchronize_seqscans &&
+								!RelationUsesLocalBuffers(rel) &&
+								bpscan->phs_nblocks > NBuffers / 4;
+	SpinLockInit(&bpscan->phs_mutex);
+	bpscan->phs_startblock = InvalidBlockNumber;
+	pg_atomic_init_u64(&bpscan->phs_nallocated, 0);
+
+	return sizeof(ParallelBlockTableScanDescData);
 }
 
 
 static void
 columnar_parallelscan_reinitialize(Relation rel, ParallelTableScanDesc pscan)
 {
-	elog(ERROR, "columnar_parallelscan_reinitialize not implemented");
+	ParallelBlockTableScanDesc bpscan = (ParallelBlockTableScanDesc) pscan;
+	pg_atomic_write_u64(&bpscan->phs_nallocated, 0);
 }
 
 
@@ -1101,10 +1113,21 @@ columnar_index_build_range_scan(Relation columnarRelation,
 	if (scan)
 	{
 		/*
-		 * Since we don't support parallel reads on columnar tables, we
-		 * should have already errored out for that, but be on the safe side.
+		 * Scan is initialized iff postgres decided to build the index using
+		 * parallel workers. In this case, we simply return for parallel
+		 * workers since we don't support parallel scan on columnar tables.
 		 */
-		ereport(ERROR, (errmsg("parallel reads on columnar are not supported")));
+		if (IsBackgroundWorker)
+		{
+			ereport(DEBUG4, (errmsg("ignoring parallel worker when building "
+									"index since parallel scan on columnar "
+									"tables is not supported")));
+			return 0;
+		}
+
+		ereport(NOTICE, (errmsg("falling back to serial index build since "
+								"parallel scan on columnar tables is not "
+								"supported")));
 	}
 
 	/*
@@ -1643,6 +1666,17 @@ static const TableAmRoutine columnar_am_methods = {
 	.scan_rescan = columnar_rescan,
 	.scan_getnextslot = columnar_getnextslot,
 
+	/*
+	 * Postgres calls following three callbacks during index builds, if it
+	 * decides to use parallel workers when building the index. On the other
+	 * hand, we don't support parallel scans on columnar tables but we also
+	 * want to fallback to serial index build. For this reason, we both skip
+	 * parallel workers in columnar_index_build_range_scan and also provide
+	 * basic implementations for those callbacks based on their corresponding
+	 * implementations in heapAM.
+	 * Note that for regular query plans, we already ignore parallel paths via
+	 * ColumnarSetRelPathlistHook.
+	 */
 	.parallelscan_estimate = columnar_parallelscan_estimate,
 	.parallelscan_initialize = columnar_parallelscan_initialize,
 	.parallelscan_reinitialize = columnar_parallelscan_reinitialize,
diff --git a/src/test/regress/expected/columnar_indexes.out b/src/test/regress/expected/columnar_indexes.out
index 6062cae34..32af88c92 100644
--- a/src/test/regress/expected/columnar_indexes.out
+++ b/src/test/regress/expected/columnar_indexes.out
@@ -384,5 +384,14 @@ INSERT INTO box_temp SELECT box(point(i, i), point(i * 2, i * 2)) FROM generate_
 CREATE TABLE brin_summarize (value int) USING columnar;
 CREATE INDEX brin_summarize_idx ON brin_summarize USING brin (value) WITH (pages_per_range=2);
 ERROR:  only btree and hash indexes are supported on columnar tables
+-- Show that we safely fallback to serial index build.
+CREATE TABLE parallel_scan_test(a int) USING columnar WITH ( parallel_workers = 2 );
+INSERT INTO parallel_scan_test SELECT i FROM generate_series(1,10) i;
+CREATE INDEX ON parallel_scan_test (a);
+NOTICE:  falling back to serial index build since parallel scan on columnar tables is not supported
+VACUUM FULL parallel_scan_test;
+NOTICE:  falling back to serial index build since parallel scan on columnar tables is not supported
+REINDEX TABLE parallel_scan_test;
+NOTICE:  falling back to serial index build since parallel scan on columnar tables is not supported
 SET client_min_messages TO WARNING;
 DROP SCHEMA columnar_indexes CASCADE;
diff --git a/src/test/regress/sql/columnar_indexes.sql b/src/test/regress/sql/columnar_indexes.sql
index 60991b56a..2b48e2a37 100644
--- a/src/test/regress/sql/columnar_indexes.sql
+++ b/src/test/regress/sql/columnar_indexes.sql
@@ -281,5 +281,12 @@ INSERT INTO box_temp SELECT box(point(i, i), point(i * 2, i * 2)) FROM generate_
 CREATE TABLE brin_summarize (value int) USING columnar;
 CREATE INDEX brin_summarize_idx ON brin_summarize USING brin (value) WITH (pages_per_range=2);
 
+-- Show that we safely fallback to serial index build.
+CREATE TABLE parallel_scan_test(a int) USING columnar WITH ( parallel_workers = 2 );
+INSERT INTO parallel_scan_test SELECT i FROM generate_series(1,10) i;
+CREATE INDEX ON parallel_scan_test (a);
+VACUUM FULL parallel_scan_test;
+REINDEX TABLE parallel_scan_test;
+
 SET client_min_messages TO WARNING;
 DROP SCHEMA columnar_indexes CASCADE;