Have the reindexer reindex more river_job indexes (#963)

Here, expand the scope of the reindexer so that beyond the pair of
indexes it reindexes right now, it also reindexes all the other ones in
the `river_job` table. This won't make a huge different for many uses of
River, but it'll make a particular difference in cases where for some
amount of time `river_job` became very large, but has since contracted
to a much more modest size. B-trees expand but never shrink, so in that
case the indexes will be left very large despite holding very little,
and the only way to get them back down to size is to rebuild them.

More reindexes will always put more load on the database, but we're
still doing that as carefully as we can in that indexes are rebuilt one
by one, and using `CONCURRENTLY`. Back in #935 we added protection so
that in case of a situation where index builds are too expensive and
start to fail, the reindexer will bail out instead of trying and failing
over and over again.

Author: brandur

CHANGELOG.md

@@ -17,6 +17,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 - Remove unecessary transactions where a single database operation will do. This reduces the number of subtransactions created which can be an operational benefit it many cases. [PR #950](https://github.com/riverqueue/river/pull/950)
 - Bring all driver tests into separate package so they don't leak dependencies. This removes dependencies from the top level `river` package that most River installations won't need, thereby reducing the transitive dependency load of most River installations. [PR #955](https://github.com/riverqueue/river/pull/955).
+- The reindexer maintenance service now reindexes all `river_job` indexes, including its primary key. This is expected to help in situations where the jobs table has in the past expanded to a very large size (which makes most indexes larger), is now a much more modest size, but has left the indexes in their expanded state. [PR #963](https://github.com/riverqueue/river/pull/963).
 - The River CLI now accepts a `--target-version` of 0 with `river migrate-down` to run all down migrations and remove all River tables (previously, -1 was used for this; -1 still works, but now 0 also works). [PR #966](https://github.com/riverqueue/river/pull/966).
 
 ### Fixed

client_test.go

@@ -6428,7 +6428,13 @@ func Test_NewClient_Defaults(t *testing.T) {
 	require.False(t, enqueuer.StaggerStartupIsDisabled())
 
 	reindexer := maintenance.GetService[*maintenance.Reindexer](client.queueMaintainer)
-	require.Equal(t, []string{"river_job_args_index", "river_job_metadata_index"}, reindexer.Config.IndexNames)
+	require.Contains(t, reindexer.Config.IndexNames, "river_job_args_index")
+	require.Contains(t, reindexer.Config.IndexNames, "river_job_kind")
+	require.Contains(t, reindexer.Config.IndexNames, "river_job_metadata_index")
+	require.Contains(t, reindexer.Config.IndexNames, "river_job_pkey")
+	require.Contains(t, reindexer.Config.IndexNames, "river_job_prioritized_fetching_index")
+	require.Contains(t, reindexer.Config.IndexNames, "river_job_state_and_finalized_at_index")
+	require.Contains(t, reindexer.Config.IndexNames, "river_job_unique_idx")
 	now := time.Now().UTC()
 	nextMidnight := time.Date(now.Year(), now.Month(), now.Day(), 0, 0, 0, 0, time.UTC).AddDate(0, 0, 1)
 	require.Equal(t, nextMidnight, reindexer.Config.ScheduleFunc(now))

internal/maintenance/reindexer.go

@@ -26,7 +26,15 @@ const (
 	ReindexerTimeoutDefault = 1 * time.Minute
 )
 
-var defaultIndexNames = []string{"river_job_args_index", "river_job_metadata_index"} //nolint:gochecknoglobals
+var defaultIndexNames = []string{ //nolint:gochecknoglobals
+	"river_job_args_index",
+	"river_job_kind",
+	"river_job_metadata_index",
+	"river_job_pkey",
+	"river_job_prioritized_fetching_index",
+	"river_job_state_and_finalized_at_index",
+	"river_job_unique_idx",
+}
 
 // Test-only properties.
 type ReindexerTestSignals struct {

internal/maintenance/reindexer_test.go

@@ -127,7 +127,7 @@ func TestReindexer(t *testing.T) {
 		require.True(t, requireReindexOne(indexName))
 	})
 
-	t.Run("ReindexesEachIndex", func(t *testing.T) {
+	t.Run("ReindexesMinimalSubsetofIndexes", func(t *testing.T) {
 		t.Parallel()
 
 		svc, bundle := setup(t)
@@ -163,6 +163,17 @@ func TestReindexer(t *testing.T) {
 		}
 	})
 
+	t.Run("ReindexesDefaultIndexes", func(t *testing.T) {
+		t.Parallel()
+
+		svc, _ := setup(t)
+
+		svc.Config.ScheduleFunc = runImmediatelyThenOnceAnHour()
+
+		require.NoError(t, svc.Start(ctx))
+		svc.TestSignals.Reindexed.WaitOrTimeout()
+	})
+
 	t.Run("ReindexDeletesArtifactsWhenCancelledWithStop", func(t *testing.T) {
 		t.Parallel()