Add cluster upgrade support by including MOFED wait nodeSelector in RDMA scenarios.

tginer · tginer · commit 914e8e624587 · 2026-02-04T12:34:30.000+01:00
Signed-off-by: Teresa Giner &lt;tginer@redhat.com&gt;
diff --git a/controllers/object_controls.go b/controllers/object_controls.go
@@ -1143,6 +1143,13 @@ func TransformDriver(obj *appsv1.DaemonSet, config *gpuv1.ClusterPolicySpec, n C
 		setContainerEnv(driverToolkitContainer, "DRIVER_CONFIG_DIGEST", configDigest)
 	}
 
+	// add nodeSelector for MOFED wait label when GPUDirect RDMA is enabled
+	if config.Driver.GPUDirectRDMA != nil && config.Driver.GPUDirectRDMA.IsEnabled() {
+		if obj.Spec.Template.Spec.NodeSelector == nil {
+			obj.Spec.Template.Spec.NodeSelector = make(map[string]string)
+		}
+		obj.Spec.Template.Spec.NodeSelector["network.nvidia.com/operator.mofed.wait"] = "false"
+	}
 	return nil
 }
 

Original file line number	Diff line number	Diff line change
`@@ -1143,6 +1143,13 @@ func TransformDriver(obj appsv1.DaemonSet, config gpuv1.ClusterPolicySpec, n C`
`1143`	`1143`	`setContainerEnv(driverToolkitContainer, "DRIVER_CONFIG_DIGEST", configDigest)`
`1144`	`1144`	`}`
`1145`	`1145`
	`1146`	`+ // add nodeSelector for MOFED wait label when GPUDirect RDMA is enabled`
	`1147`	`+ if config.Driver.GPUDirectRDMA != nil && config.Driver.GPUDirectRDMA.IsEnabled() {`
	`1148`	`+ if obj.Spec.Template.Spec.NodeSelector == nil {`
	`1149`	`+ obj.Spec.Template.Spec.NodeSelector = make(map[string]string)`
	`1150`	`+ }`
	`1151`	`+ obj.Spec.Template.Spec.NodeSelector["network.nvidia.com/operator.mofed.wait"] = "false"`
	`1152`	`+ }`
`1146`	`1153`	`return nil`
`1147`	`1154`	`}`
`1148`	`1155`