Image check improvements

charts/agent-stack-k8s/values.schema.json

@@ -491,6 +491,18 @@
             }
           }
         },
+        "default-image-pull-policy": {
+          "type": "string",
+          "description": "Configures a default image pull policy for containers that do not specify a pull policy, or containers created by the stack itself",
+          "default": "IfNotPresent",
+          "examples": ["Always", "IfNotPresent", "Never", ""]
+        },
+        "default-image-check-pull-policy": {
+          "type": "string",
+          "description": "Configures a default image pull policy for image-check init containers, used if an image pull policy is not set for the corresponding container in a podSpec or podSpecPatch",
+          "default": "",
+          "examples": ["Always", "IfNotPresent", "Never", ""]
+        },
         "pod-spec-patch": {
           "$ref": "https://kubernetesjsonschema.dev/master/_definitions.json#/definitions/io.k8s.api.core.v1.PodSpec"
         }

cmd/controller/controller.go

@@ -115,6 +115,16 @@ func AddConfigFlags(cmd *cobra.Command) {
 		config.DefaultEmptyJobGracePeriod,
 		"Duration after starting a Kubernetes job that the controller will wait before considering failing the job due to a missing pod (e.g. when the podSpec specifies a missing service account)",
 	)
+	cmd.Flags().String(
+		"default-image-pull-policy",
+		"IfNotPresent",
+		"Configures a default image pull policy for containers that do not specify a pull policy and non-init containers created by the stack itself",
+	)
+	cmd.Flags().String(
+		"default-image-check-pull-policy",
+		"",
+		"Sets a default PullPolicy for image-check init containers, used if an image pull policy is not set for the corresponding container in a podSpec or podSpecPatch",
+	)
 	cmd.Flags().Bool(
 		"prohibit-kubernetes-plugin",
 		false,

cmd/controller/controller_test.go

@@ -37,6 +37,8 @@ func TestReadAndParseConfig(t *testing.T) {
 		ClusterUUID:                  "beefcafe-abbe-baba-abba-deedcedecade",
 		ProhibitKubernetesPlugin:     true,
 		GraphQLEndpoint:              "http://graphql.buildkite.localhost/v1",
+		DefaultImagePullPolicy:       "Never",
+		DefaultImageCheckPullPolicy:  "IfNotPresent",
 
 		WorkspaceVolume: &corev1.Volume{
 			Name: "workspace-2-the-reckoning",

examples/config.yaml

@@ -11,6 +11,8 @@ job-creation-concurrency: 5
 max-in-flight: 100
 namespace: my-buildkite-ns
 org: my-buildkite-org
+default-image-pull-policy: Never
+default-image-check-pull-policy: IfNotPresent
 
 # Setting a custom GraphQL endpoint is usually only useful if you have a
 #  different instance of Buildkite itself available to run.

go.mod

@@ -6,6 +6,7 @@ require (
 	github.com/Khan/genqlient v0.7.0
 	github.com/buildkite/go-buildkite/v3 v3.13.0
 	github.com/buildkite/roko v1.2.0
+	github.com/distribution/reference v0.6.0
 	github.com/go-playground/locales v0.14.1
 	github.com/go-playground/universal-translator v0.18.1
 	github.com/go-playground/validator/v10 v10.23.0
@@ -69,7 +70,7 @@ require (
 	github.com/dustinkirkland/golang-petname v0.0.0-20240428194347-eebcea082ee0 // indirect
 	github.com/eapache/queue/v2 v2.0.0-20230407133247-75960ed334e4 // indirect
 	github.com/ebitengine/purego v0.7.1 // indirect
-	github.com/fatih/color v1.16.0 // indirect
+	github.com/fatih/color v1.18.0 // indirect
 	github.com/felixge/httpsnoop v1.0.4 // indirect
 	github.com/fxamacker/cbor/v2 v2.7.0 // indirect
 	github.com/gabriel-vasile/mimetype v1.4.3 // indirect
@@ -95,6 +96,9 @@ require (
 	github.com/mattn/go-colorable v0.1.13 // indirect
 	github.com/mattn/go-isatty v0.0.20 // indirect
 	github.com/mattn/go-runewidth v0.0.15 // indirect
+	github.com/onsi/ginkgo/v2 v2.22.0 // indirect
+	github.com/onsi/gomega v1.36.0 // indirect
+	github.com/opencontainers/go-digest v1.0.0 // indirect
 	github.com/outcaste-io/ristretto v0.2.3 // indirect
 	github.com/pkg/browser v0.0.0-20240102092130-5ac0b6a4141c // indirect
 	github.com/power-devops/perfstat v0.0.0-20220216144756-c35f1ee13d7c // indirect
@@ -149,7 +153,7 @@ require (
 	github.com/davecgh/go-spew v1.1.2-0.20180830191138-d8f796af33cc // indirect
 	github.com/denisbrodbeck/machineid v1.0.1 // indirect
 	github.com/dustin/go-humanize v1.0.1 // indirect
-	github.com/emicklei/go-restful/v3 v3.11.0 // indirect
+	github.com/emicklei/go-restful/v3 v3.12.1 // indirect
 	github.com/fsnotify/fsnotify v1.7.0 // indirect
 	github.com/go-chi/chi/v5 v5.1.0 // indirect
 	github.com/go-logr/logr v1.4.2 // indirect
@@ -194,7 +198,7 @@ require (
 	github.com/russross/blackfriday/v2 v2.1.0 // indirect
 	github.com/secure-systems-lab/go-securesystemslib v0.8.0 // indirect
 	github.com/spf13/afero v1.11.0 // indirect
-	github.com/spf13/cast v1.6.0 // indirect
+	github.com/spf13/cast v1.7.0 // indirect
 	github.com/spf13/pflag v1.0.5
 	github.com/stretchr/testify v1.10.0
 	github.com/subosito/gotenv v1.6.0 // indirect
@@ -216,7 +220,7 @@ require (
 	go.opentelemetry.io/proto/otlp v1.3.1 // indirect
 	go.uber.org/multierr v1.11.0 // indirect
 	golang.org/x/crypto v0.29.0 // indirect
-	golang.org/x/exp v0.0.0-20240613232115-7f521ea00fb8 // indirect
+	golang.org/x/exp v0.0.0-20240719175910-8a7402abbf56 // indirect
 	golang.org/x/net v0.31.0 // indirect
 	golang.org/x/oauth2 v0.24.0 // indirect
 	golang.org/x/sys v0.27.0 // indirect

go.sum

@@ -159,6 +159,8 @@ github.com/dgryski/go-rendezvous v0.0.0-20200823014737-9f7001d12a5f h1:lO4WD4F/r
 github.com/dgryski/go-rendezvous v0.0.0-20200823014737-9f7001d12a5f/go.mod h1:cuUVRXasLTGF7a8hSLbxyZXjz+1KgoB3wDUb6vlszIc=
 github.com/dgryski/trifles v0.0.0-20200323201526-dd97f9abfb48 h1:fRzb/w+pyskVMQ+UbP35JkH8yB7MYb4q/qhBarqZE6g=
 github.com/dgryski/trifles v0.0.0-20200323201526-dd97f9abfb48/go.mod h1:if7Fbed8SFyPtHLHbg49SI7NAdJiC5WIA09pe59rfAA=
+github.com/distribution/reference v0.6.0 h1:0IXCQ5g4/QMHHkarYzh5l+u8T3t73zM5QvfrDyIgxBk=
+github.com/distribution/reference v0.6.0/go.mod h1:BbU0aIcezP1/5jX/8MP0YiH4SdvB5Y4f/wlDRiLyi3E=
 github.com/dnephin/pflag v1.0.7 h1:oxONGlWxhmUct0YzKTgrpQv9AUA1wtPBn7zuSjJqptk=
 github.com/dnephin/pflag v1.0.7/go.mod h1:uxE91IoWURlOiTUIA8Mq5ZZkAv3dPUfZNaT80Zm7OQE=
 github.com/dustin/go-humanize v1.0.0/go.mod h1:HtrtbFcZ19U5GC7JDqmcUSB87Iq5E25KnS6fMYU6eOk=
@@ -170,16 +172,17 @@ github.com/eapache/queue/v2 v2.0.0-20230407133247-75960ed334e4 h1:8EXxF+tCLqaVk8
 github.com/eapache/queue/v2 v2.0.0-20230407133247-75960ed334e4/go.mod h1:I5sHm0Y0T1u5YjlyqC5GVArM7aNZRUYtTjmJ8mPJFds=
 github.com/ebitengine/purego v0.7.1 h1:6/55d26lG3o9VCZX8lping+bZcmShseiqlh2bnUDiPA=
 github.com/ebitengine/purego v0.7.1/go.mod h1:ah1In8AOtksoNK6yk5z1HTJeUkC1Ez4Wk2idgGslMwQ=
-github.com/emicklei/go-restful/v3 v3.11.0 h1:rAQeMHw1c7zTmncogyy8VvRZwtkmkZ4FxERmMY4rD+g=
-github.com/emicklei/go-restful/v3 v3.11.0/go.mod h1:6n3XBCmQQb25CM2LCACGz8ukIrRry+4bhvbpWn3mrbc=
+github.com/emicklei/go-restful/v3 v3.12.1 h1:PJMDIM/ak7btuL8Ex0iYET9hxM3CI2sjZtzpL63nKAU=
+github.com/emicklei/go-restful/v3 v3.12.1/go.mod h1:6n3XBCmQQb25CM2LCACGz8ukIrRry+4bhvbpWn3mrbc=
 github.com/envoyproxy/go-control-plane v0.9.0/go.mod h1:YTl/9mNaCwkRvm6d1a2C3ymFceY/DCBVvsKhRF0iEA4=
 github.com/envoyproxy/go-control-plane v0.9.1-0.20191026205805-5f8ba28d4473/go.mod h1:YTl/9mNaCwkRvm6d1a2C3ymFceY/DCBVvsKhRF0iEA4=
 github.com/envoyproxy/go-control-plane v0.9.4/go.mod h1:6rpuAdCZL397s3pYoYcLgu1mIlRU8Am5FuJP05cCM98=
 github.com/envoyproxy/protoc-gen-validate v0.1.0/go.mod h1:iSmxcyjqTsJpI2R4NaDN7+kN2VEUnK/pcBlmesArF7c=
 github.com/fatih/color v1.7.0/go.mod h1:Zm6kSWBoL9eyXnKyktHP6abPY2pDugNf5KwzbycvMj4=
 github.com/fatih/color v1.15.0/go.mod h1:0h5ZqXfHYED7Bhv2ZJamyIOUej9KtShiJESRwBDUSsw=
-github.com/fatih/color v1.16.0 h1:zmkK9Ngbjj+K0yRhTVONQh1p/HknKYSlNT+vZCzyokM=
 github.com/fatih/color v1.16.0/go.mod h1:fL2Sau1YI5c0pdGEVCbKQbLXB6edEj1ZgiY4NijnWvE=
+github.com/fatih/color v1.18.0 h1:S8gINlzdQ840/4pfAwic/ZE0djQEH3wM94VfqLTZcOM=
+github.com/fatih/color v1.18.0/go.mod h1:4FelSpRwEGDpQ12mAdzqdOukCy4u8WUtOY6lkT/6HfU=
 github.com/felixge/httpsnoop v1.0.4 h1:NFTV2Zj1bL4mc9sqWACXbQFVBBg2W3GPvqp8/ESS2Wg=
 github.com/felixge/httpsnoop v1.0.4/go.mod h1:m8KPJKqk1gH5J9DgRY2ASl2lWCfGKXixSwevea8zH2U=
 github.com/frankban/quicktest v1.14.6 h1:7Xjx+VpznH+oBnejlPUj8oUpdxnVs4f8XU8WnHkI4W8=
@@ -367,10 +370,12 @@ github.com/munnerz/goautoneg v0.0.0-20191010083416-a7dc8b61c822 h1:C3w9PqII01/Oq
 github.com/munnerz/goautoneg v0.0.0-20191010083416-a7dc8b61c822/go.mod h1:+n7T8mK8HuQTcFwEeznm/DIxMOiR9yIdICNftLE1DvQ=
 github.com/oleiade/reflections v1.1.0 h1:D+I/UsXQB4esMathlt0kkZRJZdUDmhv5zGi/HOwYTWo=
 github.com/oleiade/reflections v1.1.0/go.mod h1:mCxx0QseeVCHs5Um5HhJeCKVC7AwS8kO67tky4rdisA=
-github.com/onsi/ginkgo/v2 v2.21.0 h1:7rg/4f3rB88pb5obDgNZrNHrQ4e6WpjonchcpuBRnZM=
-github.com/onsi/ginkgo/v2 v2.21.0/go.mod h1:7Du3c42kxCUegi0IImZ1wUQzMBVecgIHjR1C+NkhLQo=
-github.com/onsi/gomega v1.35.1 h1:Cwbd75ZBPxFSuZ6T+rN/WCb/gOc6YgFBXLlZLhC7Ds4=
-github.com/onsi/gomega v1.35.1/go.mod h1:PvZbdDc8J6XJEpDK4HCuRBm8a6Fzp9/DmhC9C7yFlog=
+github.com/onsi/ginkgo/v2 v2.22.0 h1:Yed107/8DjTr0lKCNt7Dn8yQ6ybuDRQoMGrNFKzMfHg=
+github.com/onsi/ginkgo/v2 v2.22.0/go.mod h1:7Du3c42kxCUegi0IImZ1wUQzMBVecgIHjR1C+NkhLQo=
+github.com/onsi/gomega v1.36.0 h1:Pb12RlruUtj4XUuPUqeEWc6j5DkVVVA49Uf6YLfC95Y=
+github.com/onsi/gomega v1.36.0/go.mod h1:PvZbdDc8J6XJEpDK4HCuRBm8a6Fzp9/DmhC9C7yFlog=
+github.com/opencontainers/go-digest v1.0.0 h1:apOUWs51W5PlhuyGyz9FCeeBIOUDA/6nW8Oi/yOhh5U=
+github.com/opencontainers/go-digest v1.0.0/go.mod h1:0JzlMkj0TRzQZfJkVvzbP0HBR3IKzErnv2BNG4W4MAM=
 github.com/opentracing/opentracing-go v1.2.0 h1:uEJPy/1a5RIPAJ0Ov+OIO8OxWu77jEv+1B0VhjKrZUs=
 github.com/opentracing/opentracing-go v1.2.0/go.mod h1:GxEUsuufX4nBwe+T+Wl9TAgYrxe9dPLANfrWvHYVTgc=
 github.com/outcaste-io/ristretto v0.2.3 h1:AK4zt/fJ76kjlYObOeNwh4T3asEuaCmp26pOvUOL9w0=
@@ -448,8 +453,8 @@ github.com/spaolacci/murmur3 v1.1.0 h1:7c1g84S4BPRrfL5Xrdp6fOJ206sU9y293DDHaoy0b
 github.com/spaolacci/murmur3 v1.1.0/go.mod h1:JwIasOWyU6f++ZhiEuf87xNszmSA2myDM2Kzu9HwQUA=
 github.com/spf13/afero v1.11.0 h1:WJQKhtpdm3v2IzqG8VMqrr6Rf3UYpEF239Jy9wNepM8=
 github.com/spf13/afero v1.11.0/go.mod h1:GH9Y3pIexgf1MTIWtNGyogA5MwRIDXGUr+hbWNoBjkY=
-github.com/spf13/cast v1.6.0 h1:GEiTHELF+vaR5dhz3VqZfFSzZjYbgeKDpBxQVS4GYJ0=
-github.com/spf13/cast v1.6.0/go.mod h1:ancEpBxwJDODSW/UG4rDrAqiKolqNNh2DX3mk86cAdo=
+github.com/spf13/cast v1.7.0 h1:ntdiHjuueXFgm5nzDRdOS4yfT43P5Fnud6DH50rz/7w=
+github.com/spf13/cast v1.7.0/go.mod h1:ancEpBxwJDODSW/UG4rDrAqiKolqNNh2DX3mk86cAdo=
 github.com/spf13/cobra v1.8.1 h1:e5/vxKd/rZsfSJMUX1agtjeTDf+qv1/JdBF8gg5k9ZM=
 github.com/spf13/cobra v1.8.1/go.mod h1:wHxEcudfqmLYa8iTfL+OuZPbBZkmvliBWKIezN3kD9Y=
 github.com/spf13/pflag v1.0.5 h1:iy+VFUOCP1a+8yFto/drg2CJ5u0yRoB7fZw3DKv/JXA=
@@ -563,8 +568,8 @@ golang.org/x/crypto v0.21.0/go.mod h1:0BP7YvVV9gBbVKyeTG0Gyn+gZm94bibOW5BjDEYAOM
 golang.org/x/crypto v0.29.0 h1:L5SG1JTTXupVV3n6sUqMTeWbjAyfPwoda2DLX8J8FrQ=
 golang.org/x/crypto v0.29.0/go.mod h1:+F4F4N5hv6v38hfeYwTdx20oUvLLc+QfrE9Ax9HtgRg=
 golang.org/x/exp v0.0.0-20190121172915-509febef88a4/go.mod h1:CJ0aWSM057203Lf6IL+f9T1iT9GByDxfZKAQTCR3kQA=
-golang.org/x/exp v0.0.0-20240613232115-7f521ea00fb8 h1:yixxcjnhBmY0nkL253HFVIm0JsFHwrHdT3Yh6szTnfY=
-golang.org/x/exp v0.0.0-20240613232115-7f521ea00fb8/go.mod h1:jj3sYF3dwk5D+ghuXyeI3r5MFf+NT2An6/9dOA95KSI=
+golang.org/x/exp v0.0.0-20240719175910-8a7402abbf56 h1:2dVuKD2vS7b0QIHQbpyTISPd0LeHDbnYEryqj5Q1ug8=
+golang.org/x/exp v0.0.0-20240719175910-8a7402abbf56/go.mod h1:M4RDyNAINzryxdtnbRXRL/OHtkFuWGRjvuhBJpk2IlY=
 golang.org/x/lint v0.0.0-20181026193005-c67002cb31c3/go.mod h1:UVdnD1Gm6xHRNCYTkRU2/jEulfH38KcIWyp/GAMgvoE=
 golang.org/x/lint v0.0.0-20190227174305-5b3e6a55c961/go.mod h1:wehouNa3lNwaWXcvxsM5YxQ5yQlVC4a0KAMCusXpPoU=
 golang.org/x/lint v0.0.0-20190313153728-d0100b6bd8b3/go.mod h1:6SW0HCj/g11FgYtHlgUYUwCkIfeOF89ocIRzGO/8vkc=

internal/controller/config/config.go

@@ -64,6 +64,9 @@ type Config struct {
 	DefaultSidecarParams  *SidecarParams  `json:"default-sidecar-params"  validate:"omitempty"`
 	DefaultMetadata       Metadata        `json:"default-metadata"        validate:"omitempty"`
 
+	DefaultImagePullPolicy      corev1.PullPolicy `json:"default-image-pull-policy"       validate:"omitempty"`
+	DefaultImageCheckPullPolicy corev1.PullPolicy `json:"default-image-check-pull-policy" validate:"omitempty"`
+
 	// ProhibitKubernetesPlugin can be used to prevent alterations to the pod
 	// from the job (the kubernetes "plugin" in pipeline.yml). If enabled,
 	// jobs with a "kubernetes" plugin will fail.
@@ -120,6 +123,8 @@ func (c Config) MarshalLogObject(enc zapcore.ObjectEncoder) error {
 	if err := enc.AddReflected("default-metadata", c.DefaultMetadata); err != nil {
 		return err
 	}
+	enc.AddString("default-image-pull-policy", string(c.DefaultImagePullPolicy))
+	enc.AddString("default-image-check-pull-policy", string(c.DefaultImageCheckPullPolicy))
 	return nil
 }
 

internal/controller/controller.go

@@ -87,19 +87,21 @@ func Run(
 	// Scheduler does the complicated work of converting a Buildkite job into
 	// a pod to run that job. It talks to the k8s API to create pods.
 	sched := scheduler.New(logger.Named("scheduler"), k8sClient, scheduler.Config{
-		Namespace:              cfg.Namespace,
-		Image:                  cfg.Image,
-		AgentTokenSecretName:   cfg.AgentTokenSecret,
-		JobTTL:                 cfg.JobTTL,
-		AdditionalRedactedVars: cfg.AdditionalRedactedVars,
-		WorkspaceVolume:        cfg.WorkspaceVolume,
-		AgentConfig:            cfg.AgentConfig,
-		DefaultCheckoutParams:  cfg.DefaultCheckoutParams,
-		DefaultCommandParams:   cfg.DefaultCommandParams,
-		DefaultSidecarParams:   cfg.DefaultSidecarParams,
-		DefaultMetadata:        cfg.DefaultMetadata,
-		PodSpecPatch:           cfg.PodSpecPatch,
-		ProhibitK8sPlugin:      cfg.ProhibitKubernetesPlugin,
+		Namespace:                   cfg.Namespace,
+		Image:                       cfg.Image,
+		AgentTokenSecretName:        cfg.AgentTokenSecret,
+		JobTTL:                      cfg.JobTTL,
+		AdditionalRedactedVars:      cfg.AdditionalRedactedVars,
+		WorkspaceVolume:             cfg.WorkspaceVolume,
+		AgentConfig:                 cfg.AgentConfig,
+		DefaultCheckoutParams:       cfg.DefaultCheckoutParams,
+		DefaultCommandParams:        cfg.DefaultCommandParams,
+		DefaultSidecarParams:        cfg.DefaultSidecarParams,
+		DefaultMetadata:             cfg.DefaultMetadata,
+		DefaultImagePullPolicy:      cfg.DefaultImagePullPolicy,
+		DefaultImageCheckPullPolicy: cfg.DefaultImageCheckPullPolicy,
+		PodSpecPatch:                cfg.PodSpecPatch,
+		ProhibitK8sPlugin:           cfg.ProhibitKubernetesPlugin,
 	})
 
 	informerFactory, err := NewInformerFactory(k8sClient, cfg.Namespace, cfg.Tags)

internal/controller/scheduler/job_watcher.go

@@ -38,7 +38,7 @@ type jobWatcher struct {
 
 	// Tracks stalling jobs (jobs that have yet to create pods).
 	stallingJobsMu sync.Mutex
-	stallingJobs   map[*batchv1.Job]struct{}
+	stallingJobs   map[uuid.UUID]*batchv1.Job
 
 	// Tracks jobs that are being cleaned up (to avoid repeats).
 	ignoredJobsMu sync.RWMutex
@@ -58,7 +58,7 @@ func NewJobWatcher(logger *zap.Logger, k8sClient kubernetes.Interface, cfg *conf
 		logger:       logger,
 		k8s:          k8sClient,
 		cfg:          cfg,
-		stallingJobs: make(map[*batchv1.Job]struct{}),
+		stallingJobs: make(map[uuid.UUID]*batchv1.Job),
 		ignoredJobs:  make(map[uuid.UUID]struct{}),
 	}
 	jobsStallingGaugeFunc = func() int {
@@ -118,13 +118,14 @@ func (w *jobWatcher) OnDelete(prev any) {
 	if kjob == nil {
 		return
 	}
-	w.removeFromStalling(kjob)
 
 	jobUUID, err := jobUUIDForObject(kjob)
 	if err != nil {
 		return
 	}
 
+	w.removeFromStalling(jobUUID)
+
 	// The job is gone, so we can stop ignoring it (if it comes back).
 	w.unignoreJob(jobUUID)
 
@@ -145,10 +146,10 @@ func (w *jobWatcher) runChecks(ctx context.Context, kjob *batchv1.Job) {
 	}
 
 	if model.JobFinished(kjob) {
-		w.removeFromStalling(kjob)
+		w.removeFromStalling(jobUUID)
 		w.checkFinishedWithoutPod(ctx, log, kjob)
 	} else {
-		w.checkStalledWithoutPod(log, kjob)
+		w.checkStalledWithoutPod(log, jobUUID, kjob)
 	}
 }
 
@@ -171,7 +172,7 @@ func (w *jobWatcher) checkFinishedWithoutPod(ctx context.Context, log *zap.Logge
 	w.failJob(ctx, log, kjob, message)
 }
 
-func (w *jobWatcher) checkStalledWithoutPod(log *zap.Logger, kjob *batchv1.Job) {
+func (w *jobWatcher) checkStalledWithoutPod(log *zap.Logger, jobUUID uuid.UUID, kjob *batchv1.Job) {
 	log.Debug("Checking job for stalling without a pod")
 
 	// If the job is not finished and there is no pod, it should start one
@@ -184,7 +185,7 @@ func (w *jobWatcher) checkStalledWithoutPod(log *zap.Logger, kjob *batchv1.Job)
 	}
 	if pods > 0 {
 		// All's well with the world.
-		w.removeFromStalling(kjob)
+		w.removeFromStalling(jobUUID)
 		return
 	}
 
@@ -193,7 +194,7 @@ func (w *jobWatcher) checkStalledWithoutPod(log *zap.Logger, kjob *batchv1.Job)
 		return
 	}
 
-	w.addToStalling(kjob)
+	w.addToStalling(jobUUID, kjob)
 }
 
 func (w *jobWatcher) fetchEvents(ctx context.Context, log *zap.Logger, kjob *batchv1.Job) string {
@@ -248,16 +249,16 @@ func (w *jobWatcher) formatEvents(evlist *corev1.EventList) string {
 	return tw.Render()
 }
 
-func (w *jobWatcher) addToStalling(kjob *batchv1.Job) {
+func (w *jobWatcher) addToStalling(jobUUID uuid.UUID, kjob *batchv1.Job) {
 	w.stallingJobsMu.Lock()
 	defer w.stallingJobsMu.Unlock()
-	w.stallingJobs[kjob] = struct{}{}
+	w.stallingJobs[jobUUID] = kjob
 }
 
-func (w *jobWatcher) removeFromStalling(kjob *batchv1.Job) {
+func (w *jobWatcher) removeFromStalling(jobUUID uuid.UUID) {
 	w.stallingJobsMu.Lock()
 	defer w.stallingJobsMu.Unlock()
-	delete(w.stallingJobs, kjob)
+	delete(w.stallingJobs, jobUUID)
 }
 
 func (w *jobWatcher) stalledJobChecker(ctx context.Context) {
@@ -274,21 +275,17 @@ func (w *jobWatcher) stalledJobChecker(ctx context.Context) {
 		// Gather stalled jobs
 		var stalled []*batchv1.Job
 		w.stallingJobsMu.Lock()
-		for kjob := range w.stallingJobs {
+		for jobUUID, kjob := range w.stallingJobs {
 			if time.Since(kjob.Status.StartTime.Time) < w.cfg.EmptyJobGracePeriod {
 				continue
 			}
 
 			// ignore it from now until it is deleted
-			jobUUID, err := jobUUIDForObject(kjob)
-			if err != nil {
-				continue
-			}
 			w.ignoreJob(jobUUID)
 
 			// Move it from w.stalling into stalled
 			stalled = append(stalled, kjob)
-			delete(w.stallingJobs, kjob)
+			delete(w.stallingJobs, jobUUID)
 		}
 		w.stallingJobsMu.Unlock()
 

internal/controller/scheduler/metrics.go

@@ -133,9 +133,10 @@ var (
 // Pod watcher metrics
 
 var (
-	// Overridden to return len(jobCancelCheckers) by podWatcher.
-	jobCancelCheckerGaugeFunc      = func() int { return 0 }
-	podWatcherIgnoredJobsGaugeFunc = func() int { return 0 }
+	// Overridden by podWatcher.
+	jobCancelCheckerGaugeFunc        = func() int { return 0 }
+	podWatcherIgnoredJobsGaugeFunc   = func() int { return 0 }
+	watchingForImageFailureGaugeFunc = func() int { return 0 }
 
 	_ = promauto.NewGaugeFunc(prometheus.GaugeOpts{
 		Namespace: promNamespace,
@@ -149,6 +150,12 @@ var (
 		Name:      "num_ignored_jobs",
 		Help:      "Current count of jobs ignored for podWatcher checks",
 	}, func() float64 { return float64(podWatcherIgnoredJobsGaugeFunc()) })
+	_ = promauto.NewGaugeFunc(prometheus.GaugeOpts{
+		Namespace: promNamespace,
+		Subsystem: "pod_watcher",
+		Name:      "num_watching_for_image_failure",
+		Help:      "Current count of pods being watched for potential image-related failures",
+	}, func() float64 { return float64(watchingForImageFailureGaugeFunc()) })
 
 	podWatcherOnAddEventCounter = promauto.NewCounter(prometheus.CounterOpts{
 		Namespace: promNamespace,