fix: add -scheduler.idleSleep flag to prevent EC integration test hang

The new 17-minute default idle sleep caused TestEcEndToEnd to hang
because the scheduler would not re-check for work frequently enough
after the first iteration found nothing.

Add -scheduler.idleSleep CLI flag (in seconds) to configure the idle
sleep duration. The EC integration test now passes -scheduler.idleSleep=2
so detection runs every 2 seconds when idle.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
This commit is contained in:
Chris Lu
2026-03-03 19:05:56 -08:00
co-authored by Claude Opus 4.6
parent 6289beb8f5
commit e0cccae794
3 changed files with 13 additions and 5 deletions
@@ -135,7 +135,7 @@ func ensureEnvironment(t *testing.T) {
// 6. Start Admin
os.RemoveAll(filepath.Join("tmp", "admin"))
os.MkdirAll(filepath.Join("tmp", "admin"), 0755)
startWeed(t, "admin", "admin", "-master=localhost:9333", "-port=23646", "-dataDir=./tmp/admin")
startWeed(t, "admin", "admin", "-master=localhost:9333", "-port=23646", "-dataDir=./tmp/admin", "-scheduler.idleSleep=2")
waitForUrl(t, AdminUrl+"/health", 60)
t.Log("Environment started successfully")
+5 -3
View File
@@ -118,7 +118,7 @@ type AdminServer struct {
// Type definitions moved to types.go
func NewAdminServer(masters string, templateFS http.FileSystem, dataDir string, icebergPort int) *AdminServer {
func NewAdminServer(masters string, templateFS http.FileSystem, dataDir string, icebergPort int, idleSleepDuration time.Duration) *AdminServer {
grpcDialOption := security.LoadClientTLS(util.GetViper(), "grpc.admin")
// Create master client with multiple master support
@@ -229,7 +229,8 @@ func NewAdminServer(masters string, templateFS http.FileSystem, dataDir string,
}
plugin, err := adminplugin.New(adminplugin.Options{
DataDir: dataDir,
DataDir: dataDir,
IdleSleepDuration: idleSleepDuration,
ClusterContextProvider: func(_ context.Context) (*plugin_pb.ClusterContext, error) {
return server.buildDefaultPluginClusterContext(), nil
},
@@ -238,7 +239,8 @@ func NewAdminServer(masters string, templateFS http.FileSystem, dataDir string,
if err != nil && dataDir != "" {
glog.Warningf("Failed to initialize plugin with dataDir=%q: %v. Falling back to in-memory plugin state.", dataDir, err)
plugin, err = adminplugin.New(adminplugin.Options{
DataDir: "",
DataDir: "",
IdleSleepDuration: idleSleepDuration,
ClusterContextProvider: func(_ context.Context) (*plugin_pb.ClusterContext, error) {
return server.buildDefaultPluginClusterContext(), nil
},
+7 -1
View File
@@ -45,6 +45,7 @@ type AdminOptions struct {
readOnlyPassword *string
dataDir *string
icebergPort *int
idleSleepSeconds *int
}
func init() {
@@ -60,6 +61,7 @@ func init() {
a.readOnlyUser = cmdAdmin.Flag.String("readOnlyUser", "", "read-only user username (optional, for view-only access)")
a.readOnlyPassword = cmdAdmin.Flag.String("readOnlyPassword", "", "read-only user password (optional, for view-only access; requires adminPassword to be set)")
a.icebergPort = cmdAdmin.Flag.Int("iceberg.port", 8181, "Iceberg REST Catalog port (0 to hide in UI)")
a.idleSleepSeconds = cmdAdmin.Flag.Int("scheduler.idleSleep", 0, "scheduler idle sleep in seconds between iterations when no work is found (0 = default 17 minutes)")
}
var cmdAdmin = &Command{
@@ -290,7 +292,11 @@ func startAdminServer(ctx context.Context, options AdminOptions, enableUI bool,
}
// Create admin server (plugin is always enabled)
adminServer := dash.NewAdminServer(*options.master, nil, dataDir, icebergPort)
var idleSleep time.Duration
if options.idleSleepSeconds != nil && *options.idleSleepSeconds > 0 {
idleSleep = time.Duration(*options.idleSleepSeconds) * time.Second
}
adminServer := dash.NewAdminServer(*options.master, nil, dataDir, icebergPort, idleSleep)
// Show discovered filers
filers := adminServer.GetAllFilers()