bzp2010 commented on code in PR #2856:
URL:
https://github.com/apache/apisix-ingress-controller/pull/2856#discussion_r3988658560
##########
internal/provider/apisix/status.go:
##########
@@ -42,32 +55,101 @@ import (
func (d *apisixProvider) handleStatusUpdate(statusUpdateMap
map[types.NamespacedNameKind][]string) {
// Mark all resources in the current failure set as failed.
for nnk, msgs := range statusUpdateMap {
- d.updateStatus(nnk, cutils.NewConditionTypeAccepted(
- apiv2.ConditionReasonSyncFailed,
- false,
- 0,
- strings.Join(msgs, "; "),
- ))
+ d.updateStatus(nnk, failureCondition(nnk, strings.Join(msgs, ";
")))
}
// Mark resources that exist only in the previous failure history as
successful.
for nnk := range d.statusUpdateMap {
if _, ok := statusUpdateMap[nnk]; !ok {
- d.updateStatus(nnk, cutils.NewConditionTypeAccepted(
- apiv2.ConditionReasonAccepted,
- true,
- 0,
- "",
- ))
+ d.updateStatus(nnk, successCondition(nnk))
+ if nnk.Kind == types.KindGatewayProxy {
+ d.recordGatewayProxyRecoveredEvent(nnk)
+ }
}
}
// Update the failure history with the current failure set.
d.statusUpdateMap = statusUpdateMap
}
+// failureCondition and successCondition pick which condition a
NamespacedNameKind
+// gets: GatewayProxyConditionDataPlaneAvailable for a GatewayProxy, the
existing
+// Accepted/SyncFailed condition for everything else.
+func failureCondition(nnk types.NamespacedNameKind, msg string)
metav1.Condition {
+ if nnk.Kind == types.KindGatewayProxy {
+ return newGatewayProxyDataPlaneAvailableCondition(false,
GatewayProxyReasonDataPlaneInstanceUnavailable, msg)
+ }
+ return cutils.NewConditionTypeAccepted(apiv2.ConditionReasonSyncFailed,
false, 0, msg)
+}
+
+func successCondition(nnk types.NamespacedNameKind) metav1.Condition {
+ if nnk.Kind == types.KindGatewayProxy {
+ return newGatewayProxyDataPlaneAvailableCondition(true,
GatewayProxyReasonDataPlaneAvailable, "")
+ }
+ return cutils.NewConditionTypeAccepted(apiv2.ConditionReasonAccepted,
true, 0, "")
+}
+
+func newGatewayProxyDataPlaneAvailableCondition(available bool, reason, msg
string) metav1.Condition {
+ conditionStatus := metav1.ConditionFalse
+ if available {
+ conditionStatus = metav1.ConditionTrue
+ }
+ return metav1.Condition{
+ Type: GatewayProxyConditionDataPlaneAvailable,
+ Status: conditionStatus,
+ LastTransitionTime: metav1.Now(),
+ Reason: reason,
+ Message: cutils.TruncateConditionMessage(msg),
+ }
+}
+
+// recordGatewayProxyRecoveredEvent fires once a GatewayProxy leaves the
failure
+// history entirely, i.e. every instance took the last sync.
+func (d *apisixProvider) recordGatewayProxyRecoveredEvent(nnk
types.NamespacedNameKind) {
+ if d.EventRecorder == nil {
+ return
+ }
+ gatewayProxy := &apiv1alpha1.GatewayProxy{
+ ObjectMeta: metav1.ObjectMeta{Name: nnk.Name, Namespace:
nnk.Namespace},
+ }
+ d.EventRecorder.Event(gatewayProxy, corev1.EventTypeNormal,
GatewayProxyReasonDataPlaneAvailable,
+ "every data plane instance took the last sync")
+}
+
+// recordFailedEndpointEvents fires one Warning event per failed
EndpointStatus entry,
+// so each instance's own failure history (when it started, how often) is
visible on
+// its own, not folded into everyone else's.
+func (d *apisixProvider) recordFailedEndpointEvents(nnk
types.NamespacedNameKind, endpoints []adctypes.EndpointStatus) {
+ if d.EventRecorder == nil {
+ return
+ }
+ gatewayProxy := &apiv1alpha1.GatewayProxy{
+ ObjectMeta: metav1.ObjectMeta{Name: nnk.Name, Namespace:
nnk.Namespace},
+ }
+ for _, ep := range endpoints {
+ if ep.Success {
+ continue
+ }
+ d.EventRecorder.Event(gatewayProxy, corev1.EventTypeWarning,
GatewayProxyReasonDataPlaneInstanceUnavailable,
+ fmt.Sprintf("%s: %s", ep.Server, ep.Reason))
+ }
+}
+
//nolint:gocyclo
func (d *apisixProvider) updateStatus(nnk types.NamespacedNameKind, condition
metav1.Condition) {
switch nnk.Kind {
+ case types.KindGatewayProxy:
+ // Unlike the route kinds below, the condition lands on the
GatewayProxy's own
+ // top-level Status.Conditions, not on a per-parent entry.
+ d.updater.Update(status.Update{
Review Comment:
Confirmed `statusEqual` already special-cases GatewayProxy (see
`internal/controller/status/updater.go`), so writes there dedupe the same way
every other kind's do. No further change needed for this one.
##########
internal/provider/apisix/status.go:
##########
@@ -42,32 +55,101 @@ import (
func (d *apisixProvider) handleStatusUpdate(statusUpdateMap
map[types.NamespacedNameKind][]string) {
// Mark all resources in the current failure set as failed.
for nnk, msgs := range statusUpdateMap {
- d.updateStatus(nnk, cutils.NewConditionTypeAccepted(
- apiv2.ConditionReasonSyncFailed,
- false,
- 0,
- strings.Join(msgs, "; "),
- ))
+ d.updateStatus(nnk, failureCondition(nnk, strings.Join(msgs, ";
")))
}
// Mark resources that exist only in the previous failure history as
successful.
for nnk := range d.statusUpdateMap {
if _, ok := statusUpdateMap[nnk]; !ok {
- d.updateStatus(nnk, cutils.NewConditionTypeAccepted(
- apiv2.ConditionReasonAccepted,
- true,
- 0,
- "",
- ))
+ d.updateStatus(nnk, successCondition(nnk))
Review Comment:
Reworked the whole status pipeline: GatewayProxy's DataPlaneAvailable is now
recomputed from scratch every sync round straight from that round's actual
result, never diffed against remembered history. A round with no error writes
True, any error writes False — recovery is never inferred from a key merely
leaving a failure set. This makes the class of bug you describe structurally
impossible rather than patching the specific case. See
`updateStatusFromSyncResults`/`classifySyncResult` in
`internal/provider/apisix/status.go`.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
