test(e2e): add multi-GPU MIG profile coverage

[karenychen]

What this PR does / why we need it:

Adds an Ubuntu 24.04 E2E scenario that applies the MIG2g profile to a two-GPU Standard_NC48ads_A100_v4 VM and validates the full VM-wide result:

• the NVIDIA device plugin service is running
• MIG mode is enabled on both A100 GPUs
• 6 MIG 2g.20gb instances are created
• the node advertises nvidia.com/gpu=6
• a MIG workload can be scheduled and run successfully

The shared MIG validators now assert exact GPU and MIG-instance counts, so the test cannot pass after partitioning only one card. Existing single-GPU MIG scenarios were updated to use the same exact-count validation.

Which issue(s) this PR fixes:

Fixes #

Testing:

• cd e2e && go test ./... -run "^$" -count=1
• Local E2E: Test_Ubuntu2404_NvidiaDevicePluginRunning_MIG_MultiGPU passed with 2 GPUs, 6 MIG 2g.20gb instances, and a schedulable GPU workload
• PR GPU E2E build 173548093: Test_Ubuntu2404_NvidiaDevicePluginRunning_MIG_MultiGPU passed in 1487.40s
• git diff --check

[Copilot]

Pull request overview

This PR expands GPU managed-experience E2E coverage for NVIDIA MIG on Ubuntu 24.04 by adding a multi-GPU scenario and tightening the shared MIG validators to assert exact, VM-wide expected results (rather than allowing partial/one-GPU partitioning to pass).

Changes:

• Updated MIG validators to require exact per-GPU MIG-mode enablement and exact MIG-instance counts.
• Updated existing single-GPU MIG scenarios to use the new exact-count validators.
• Added a new Ubuntu 24.04 multi-GPU MIG scenario.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 2 comments.

File	Description
e2e/validators.go	Makes MIG validation stricter by checking per-GPU status and exact instance counts.
e2e/scenario_gpu_managed_experience_test.go	Updates existing MIG tests to use strict validators and adds a new multi-GPU MIG scenario.

Comments suppressed due to low confidence (1)

e2e/scenario_gpu_managed_experience_test.go:656

• The PR description calls out validating MIG 2g.10gb instances, but this scenario is currently asserting "MIG 2g.20gb". If the intent is ND96asr_v4 (A100 40GB), the MIG2g profile corresponds to 2g.10gb, so the expected profile string here should match that.

				ValidateMIGInstancesCreated(ctx, s, "MIG 2g.20gb", totalMIGInstances)

[Copilot]

Pull request overview

Copilot reviewed 2 out of 2 changed files in this pull request and generated no new comments.

[github-actions]

Windows Unit Test Results

  3 files   11 suites   49s ⏱️
381 tests 381 ✅ 0 💤 0 ❌
384 runs  384 ✅ 0 💤 0 ❌

Results for commit 4d63844.

♻️ This comment has been updated with latest results.

[Copilot]

Pull request overview

Copilot reviewed 2 out of 2 changed files in this pull request and generated no new comments.

Comments suppressed due to low confidence (1)

e2e/validators.go:2627

• The failure message passed to execScriptOnVMForScenarioValidateExitCode is misleading now that the script only queries MIG mode (it will still exit 0 when MIG is "Disabled"). This message should describe the command failure (querying MIG mode), while the subsequent assertions handle the enabled/disabled validation.

	command := []string{
		"set -ex",
		"sudo nvidia-smi --query-gpu=mig.mode.current --format=csv,noheader",
	}
	execResult := execScriptOnVMForScenarioValidateExitCode(ctx, s, strings.Join(command, "\n"), 0, "MIG mode is not enabled")

[Copilot]

Pull request overview

Copilot reviewed 2 out of 2 changed files in this pull request and generated no new comments.

Comments suppressed due to low confidence (1)

e2e/scenario_gpu_managed_experience_test.go:627

• The new multi-GPU A100 MIG scenario doesn’t set Location, while the other A100/MIG scenarios in this file pin to westus2. Leaving it unset can cause this scenario to run in a region without capacity for Standard_NC48ads_A100_v4, increasing flakiness and time spent searching for capacity.

		Description:      "Tests that a MIG profile is applied to every GPU on an Ubuntu 24.04 multi-GPU VM",
		K8sSystemPoolSKU: "Standard_D2s_v3",
		Tags: Tags{
			GPU: true,
		},