kubernetes_management: add large_scale_provisioning scenario

.pre-commit-config.yaml

@@ -0,0 +1,20 @@
+repos:
+  - repo: local
+    hooks:
+      - id: pyink
+        name: pyink
+        entry: pyink --pyink-indentation 2 --pyink-use-majority-quotes --unstable --line-length=80
+        language: system
+        types: [python]
+      - id: lint-diffs
+        name: lint-diffs (changed lines only)
+        entry: bash -c 'git diff -U0 upstream/master -- "$@" | lint-diffs'
+        language: system
+        types: [python]
+        pass_filenames: true
+      - id: no-print
+        name: no print statements (non-test files)
+        entry: bash -c 'grep -n "^\s*print(" "$@" && exit 1 || exit 0'
+        language: system
+        types: [python]
+        exclude: "(^tests/|^tools/|^perfkitbenchmarker/scripts/)"

apply_batchB.py

@@ -0,0 +1,180 @@
+#!/usr/bin/env python3
+"""Batch B: relocate the node-pool sweep (hubatish 242 / 275 / 292).
+
+Behavioral change requested by review:
+  - Don't sweep at the START of Run (cluster should start clean).
+  - Each scenario cleans up after itself: sweep AFTER each scenario so the
+    next one starts clean, rather than the next scenario fixing the mess.
+  - Cleanup() performs the same sweep (reviewer: "cleanup doesn't call this").
+  - Rename _CleanStartSweep -> _SweepNodePools (no longer start-specific).
+
+Run on top of Batch A (commit b0c04ec6d). Idempotent: [ok]/[SKIP] per edit.
+"""
+
+import pathlib
+
+BENCH = pathlib.Path(
+    "perfkitbenchmarker/linux_benchmarks/kubernetes_management_benchmark.py"
+)
+TEST = pathlib.Path(
+    "tests/linux_benchmarks/kubernetes_management_benchmark_test.py"
+)
+
+
+def edit(path, old, new, label, required=True):
+  text = path.read_text()
+  if new in text and old not in text:
+    print(f"[SKIP] {label} (already applied)")
+    return
+  if old not in text:
+    if required:
+      raise SystemExit(f"[FAIL] anchor not found: {label}\n---\n{old[:160]}")
+    print(f"[SKIP] {label} (anchor absent, optional)")
+    return
+  if text.count(old) != 1 and required:
+    raise SystemExit(f"[FAIL] anchor not unique ({text.count(old)}x): {label}")
+  path.write_text(text.replace(old, new, 1))
+  print(f"[ok]   {label}")
+
+
+def replace_all(path, old, new, label):
+  text = path.read_text()
+  if old not in text:
+    print(f"[SKIP] {label} (absent)")
+    return
+  n = text.count(old)
+  path.write_text(text.replace(old, new))
+  print(f"[ok]   {label} ({n}x)")
+
+
+# ───────────────────────── BENCH: rename helper ────────────────────────────
+edit(
+    BENCH,
+    "def _CleanStartSweep(cluster: kubernetes_cluster.KubernetesCluster) ->"
+    ' None:\n  """Deletes any stale pkbm* node pools so each run starts clean'
+    ' (spec C.2)."""\n  stale = [n for n in cluster.GetNodePoolNames() if'
+    ' n.startswith(_PREFIX)]\n  if not stale:\n    logging.info("CleanStart: no'
+    ' stale pools found — clean start confirmed.")\n    return\n '
+    ' logging.info("CleanStart: deleting %d stale pools: %s", len(stale),'
+    " stale)\n  background_tasks.RunThreaded(cluster.DeleteNodePool, stale)",
+    "def _SweepNodePools(cluster: kubernetes_cluster.KubernetesCluster) ->"
+    ' None:\n  """Deletes all pkbm* node pools, blocking until each delete'
+    " completes.\n\n  Called after each scenario so the next one starts from a"
+    " clean cluster,\n  and from Cleanup() as a final best-effort teardown.\n "
+    ' """\n  pools = [n for n in cluster.GetNodePoolNames() if'
+    ' n.startswith(_PREFIX)]\n  if not pools:\n    logging.info("Sweep: no'
+    ' pkbm* pools present — cluster is clean.")\n    return\n '
+    ' logging.info("Sweep: deleting %d pkbm* pools: %s", len(pools), pools)\n '
+    " background_tasks.RunThreaded(cluster.DeleteNodePool, pools)",
+    "rename _CleanStartSweep -> _SweepNodePools (242)",
+)
+
+# ───────────────────────── BENCH: Run — drop start sweep ───────────────────
+edit(
+    BENCH,
+    "  assert isinstance(cluster, kubernetes_cluster.KubernetesCluster)\n"
+    "\n"
+    "  # Spec C.2: start clean.\n"
+    "  _CleanStartSweep(cluster)\n"
+    "\n"
+    "  # Resolve the initial node-pool version once",
+    "  assert isinstance(cluster, kubernetes_cluster.KubernetesCluster)\n"
+    "\n"
+    "  # Resolve the initial node-pool version once",
+    "Run: remove start-of-run sweep (242)",
+)
+
+# ───────────────────────── BENCH: Run — per-scenario sweep ─────────────────
+edit(
+    BENCH,
+    '  if "concurrent_node_pool_ops" in scenarios:\n    samples +='
+    " _RunConcurrentNodePoolOps(cluster, initial)\n  if"
+    ' "overlapping_cluster_update" in scenarios:\n    samples +='
+    " _RunOverlappingClusterUpdate(cluster, initial)\n  if"
+    ' "large_scale_provisioning" in scenarios:\n    # Stale pools from earlier'
+    " scenarios may still be in Deleting state and\n    # count toward AKS's"
+    " 100-pool cluster limit; sweep before the scale work\n    # so we don't"
+    " hit MaxAgentPoolCountReached mid-run.\n    _CleanStartSweep(cluster)\n   "
+    " samples += _SweepScales(cluster, initial)\n",
+    '  if "concurrent_node_pool_ops" in scenarios:\n'
+    "    samples += _RunConcurrentNodePoolOps(cluster, initial)\n"
+    "    # Each scenario leaves the cluster clean for the next one.\n"
+    "    _SweepNodePools(cluster)\n"
+    '  if "overlapping_cluster_update" in scenarios:\n'
+    "    samples += _RunOverlappingClusterUpdate(cluster, initial)\n"
+    "    _SweepNodePools(cluster)\n"
+    '  if "large_scale_provisioning" in scenarios:\n'
+    "    samples += _SweepScales(cluster, initial)\n"
+    "    _SweepNodePools(cluster)\n",
+    "Run: per-scenario sweep instead of start/pre-scale (275)",
+)
+
+# ───────────────────────── BENCH: Cleanup uses the sweep ───────────────────
+edit(
+    BENCH,
+    '  kubectl.RunKubectlCommand(\n      ["delete", "pod", _SLEEP_POD_NAME,'
+    ' "--ignore-not-found"],\n      raise_on_failure=False,\n  )\n  leftover ='
+    " [n for n in cluster.GetNodePoolNames() if n.startswith(_PREFIX)]\n  if"
+    ' not leftover:\n    return\n  logging.info("Cleanup: deleting %d leftover'
+    ' node pools", len(leftover))\n '
+    " background_tasks.RunThreaded(cluster.DeleteNodePool, leftover)",
+    "  kubectl.RunKubectlCommand(\n"
+    '      ["delete", "pod", _SLEEP_POD_NAME, "--ignore-not-found"],\n'
+    "      raise_on_failure=False,\n"
+    "  )\n"
+    "  # Final teardown reuses the same sweep the scenarios use.\n"
+    "  _SweepNodePools(cluster)",
+    "Cleanup: delegate to _SweepNodePools (242)",
+)
+
+print("\nBENCH edits complete.\n")
+
+# ════════════════════════ TEST FILE ════════════════════════════════════════
+
+# Rename the helper everywhere in tests.
+replace_all(
+    TEST,
+    "_CleanStartSweep",
+    "_SweepNodePools",
+    "test: rename _CleanStartSweep -> _SweepNodePools",
+)
+
+# Rename the dedicated test class + its docstring.
+edit(
+    TEST,
+    "class CleanStartSweepTest(pkb_common_test_case.PkbCommonTestCase):",
+    "class SweepNodePoolsTest(pkb_common_test_case.PkbCommonTestCase):",
+    "test: rename CleanStartSweepTest class",
+    required=False,
+)
+edit(
+    TEST,
+    "  def testCleanStartSweepRaisesOnGetNodePoolNamesException(self):",
+    "  def testSweepRaisesOnGetNodePoolNamesException(self):",
+    "test: rename raises-on-exception test method",
+    required=False,
+)
+
+# testRunCallsCleanStartSweep: behavior changed. The sweep now runs AFTER each
+# scenario (3 scenarios default-on -> 3 sweeps), not at start + pre-scale (2).
+edit(
+    TEST,
+    '  def testRunCallsCleanStartSweep(self):\n    """Tests that Run invokes'
+    ' _SweepNodePools before executing scenarios."""',
+    "  def testRunSweepsAfterEachScenario(self):\n"
+    '    """Run sweeps node pools after each scenario that executes."""',
+    "test: rename+repurpose testRunCallsCleanStartSweep",
+    required=False,
+)
+edit(
+    TEST,
+    "      kubernetes_management_benchmark.Run(bm_spec)\n"
+    "    self.assertEqual(mock_clean.call_count, 2)",
+    "      kubernetes_management_benchmark.Run(bm_spec)\n"
+    "    # All three scenarios run by default -> one sweep after each.\n"
+    "    self.assertEqual(mock_clean.call_count, 3)",
+    "test: sweep count 2 -> 3 (per-scenario)",
+    required=False,
+)
+
+print("\nTEST edits complete.\n")