test(metrics): Add new test to collect Firecracker vhost-user metrics

Add a test to boot a VM with vhost-user as a scratch device.
Resize the vhost-user scratch drive to have config change notifications
and collect vhost-user metrics emitted by Firecracker for this test.
Export the metrics to CloudWatch.
We use a new file in performance directory for this test becuase:
1. Tests doesn't need to be ABfied so can't be part of test_block_ab
   and the Performance pipeline.
2. By running in nightly pipeline keeping the test in functional
   directory, there will be lot of other tests running in parallel
   resulting in noisy metrics.
3. Nightly pipeline runs performance tests in a decidated BK agent,
   with nothing else running on the host, and with no parallelism so
   it makes sense to move to performance directory but keep it out of
   AB tests.

Signed-off-by: Sudan Landge <[email protected]>

Author: Sudan Landge

tests/integration_tests/performance/test_vhost_user_metrics.py

@@ -0,0 +1,87 @@
+# Copyright 2023 Amazon.com, Inc. or its affiliates. All Rights Reserved.
+# SPDX-License-Identifier: Apache-2.0
+"""Tests to collect Firecracker metrics for vhost-user devices."""
+
+import time
+
+import pytest
+
+import host_tools.drive as drive_tools
+from framework.utils_drive import resize_vhost_user_drive, spawn_vhost_user_backend
+
+
+@pytest.mark.parametrize("vcpu_count", [1, 2], ids=["1vcpu", "2vcpu"])
+def test_vhost_user_block_metrics(
+    microvm_factory, guest_kernel, rootfs_ubuntu_22, vcpu_count, metrics
+):
+    """
+    This test tries to boot a VM with vhost-user-block
+    as a scratch device, resize the vhost-user scratch drive to have
+    config change notifications, collects and then uploads the related
+    vhost-user FirecrackerMetrics to Cloudwatch.
+    Having vhost-user as root device vs a scratch should not impact metrics, however,
+    we choose to have it as a scratch device because we are interested in config change
+    metrics which we cannot extract when vhost-user is root device
+    (read only rootfs won't have a config change).
+    """
+    orig_size = 10  # MB
+    # Picked from test_config_change assuming that the intention is to change size from
+    # low->high->low->high and so the numbers are not in monotonic sequence.
+    new_sizes = [20, 10, 30]  # MB
+    vhost_user_socket = "/vub.socket"
+
+    vm = microvm_factory.build(guest_kernel, rootfs_ubuntu_22, monitor_memory=False)
+    vm.spawn(log_level="Info")
+    vm.basic_config(vcpu_count=vcpu_count)
+
+    # Add a block device to test resizing.
+    fs = drive_tools.FilesystemFile(size=orig_size)
+    _backend = spawn_vhost_user_backend(vm, fs.path, vhost_user_socket)
+    vm.add_vhost_user_drive("scratch", vhost_user_socket)
+    vm.start()
+
+    # vhost-user-block is activated during boot but it takes a while so we wait.
+    # 300msec picked by the limited number of experiments tried to see how long
+    # it takes to get the activate_time_us metrics.
+    time.sleep(0.3)
+
+    metrics.set_dimensions(
+        {
+            "performance_test": "vhost_user_block_metrics",
+            "io_engine": "vhost-user",
+            **vm.dimensions,
+        }
+    )
+    fc_metrics = vm.flush_metrics()
+    assert 0 == fc_metrics["vhost_user_block_scratch"]["activate_fails"]
+    assert fc_metrics["vhost_user_block_scratch"]["init_time_us"]
+    assert fc_metrics["vhost_user_block_scratch"]["activate_time_us"]
+
+    metrics.put_metric(
+        "init_time_us",
+        fc_metrics["vhost_user_block_scratch"]["init_time_us"],
+        unit="Microseconds",
+    )
+    metrics.put_metric(
+        "activate_time_us",
+        fc_metrics["vhost_user_block_scratch"]["activate_time_us"],
+        unit="Microseconds",
+    )
+
+    for new_size in new_sizes:
+        # Instruct the backend to resize the device.
+        # It will both resize the file and update its device config.
+        resize_vhost_user_drive(vm, new_size)
+
+        # Instruct Firecracker to reread device config and notify
+        # the guest of a config change.
+        vm.patch_drive("scratch")
+
+        fc_metrics = vm.flush_metrics()
+        assert 0 == fc_metrics["vhost_user_block_scratch"]["cfg_fails"]
+        assert fc_metrics["vhost_user_block_scratch"]["config_change_time_us"]
+        metrics.put_metric(
+            "config_change_time_us",
+            fc_metrics["vhost_user_block_scratch"]["config_change_time_us"],
+            unit="Microseconds",
+        )