Update QA tests to check disk and NIC names
[ganeti-local] / qa / qa_cluster.py
index 410196b..c4de57a 100644 (file)
@@ -35,6 +35,7 @@ from ganeti import pathutils
 import qa_config
 import qa_utils
 import qa_error
+import qa_instance
 
 from qa_utils import AssertEqual, AssertCommand, GetCommandOutput
 
@@ -63,53 +64,52 @@ def _CheckFileOnAllNodes(filename, content):
     AssertEqual(qa_utils.GetCommandOutput(node.primary, cmd), content)
 
 
-# "gnt-cluster info" fields
-_CIFIELD_RE = re.compile(r"^[-\s]*(?P<field>[^\s:]+):\s*(?P<value>\S.*)$")
+def _GetClusterField(field_path):
+  """Get the value of a cluster field.
 
-
-def _GetBoolClusterField(field):
-  """Get the Boolean value of a cluster field.
-
-  This function currently assumes that the field name is unique in the cluster
-  configuration. An assertion checks this assumption.
-
-  @type field: string
-  @param field: Name of the field
-  @rtype: bool
-  @return: The effective value of the field
+  @type field_path: list of strings
+  @param field_path: Names of the groups/fields to navigate to get the desired
+      value, e.g. C{["Default node parameters", "oob_program"]}
+  @return: The effective value of the field (the actual type depends on the
+      chosen field)
 
   """
-  master = qa_config.GetMasterNode()
-  infocmd = "gnt-cluster info"
-  info_out = qa_utils.GetCommandOutput(master.primary, infocmd)
-  ret = None
-  for l in info_out.splitlines():
-    m = _CIFIELD_RE.match(l)
-    # FIXME: There should be a way to specify a field through a hierarchy
-    if m and m.group("field") == field:
-      # Make sure that ignoring the hierarchy doesn't cause a double match
-      assert ret is None
-      ret = (m.group("value").lower() == "true")
-  if ret is not None:
-    return ret
-  raise qa_error.Error("Field not found in cluster configuration: %s" % field)
+  assert isinstance(field_path, list)
+  assert field_path
+  ret = qa_utils.GetObjectInfo(["gnt-cluster", "info"])
+  for key in field_path:
+    ret = ret[key]
+  return ret
 
 
 # Cluster-verify errors (date, "ERROR", then error code)
-_CVERROR_RE = re.compile(r"^[\w\s:]+\s+- ERROR:([A-Z0-9_-]+):")
+_CVERROR_RE = re.compile(r"^[\w\s:]+\s+- (ERROR|WARNING):([A-Z0-9_-]+):")
 
 
 def _GetCVErrorCodes(cvout):
-  ret = set()
+  errs = set()
+  warns = set()
   for l in cvout.splitlines():
     m = _CVERROR_RE.match(l)
     if m:
-      ecode = m.group(1)
-      ret.add(ecode)
-  return ret
+      etype = m.group(1)
+      ecode = m.group(2)
+      if etype == "ERROR":
+        errs.add(ecode)
+      elif etype == "WARNING":
+        warns.add(ecode)
+  return (errs, warns)
+
 
+def _CheckVerifyErrors(actual, expected, etype):
+  exp_codes = compat.UniqueFrozenset(e for (_, e, _) in expected)
+  if not actual.issuperset(exp_codes):
+    missing = exp_codes.difference(actual)
+    raise qa_error.Error("Cluster-verify didn't return these expected"
+                         " %ss: %s" % (etype, utils.CommaJoin(missing)))
 
-def AssertClusterVerify(fail=False, errors=None):
+
+def AssertClusterVerify(fail=False, errors=None, warnings=None):
   """Run cluster-verify and check the result
 
   @type fail: bool
@@ -118,19 +118,20 @@ def AssertClusterVerify(fail=False, errors=None):
   @param errors: List of CV_XXX errors that are expected; if specified, all the
       errors listed must appear in cluster-verify output. A non-empty value
       implies C{fail=True}.
+  @type warnings: list of tuples
+  @param warnings: Same as C{errors} but for warnings.
 
   """
   cvcmd = "gnt-cluster verify"
   mnode = qa_config.GetMasterNode()
-  if errors:
+  if errors or warnings:
     cvout = GetCommandOutput(mnode.primary, cvcmd + " --error-codes",
-                             fail=True)
-    actual = _GetCVErrorCodes(cvout)
-    expected = compat.UniqueFrozenset(e for (_, e, _) in errors)
-    if not actual.issuperset(expected):
-      missing = expected.difference(actual)
-      raise qa_error.Error("Cluster-verify didn't return these expected"
-                           " errors: %s" % utils.CommaJoin(missing))
+                             fail=(fail or errors))
+    (act_errs, act_warns) = _GetCVErrorCodes(cvout)
+    if errors:
+      _CheckVerifyErrors(act_errs, errors, "error")
+    if warnings:
+      _CheckVerifyErrors(act_warns, warnings, "warning")
   else:
     AssertCommand(cvcmd, fail=fail, node=mnode)
 
@@ -153,7 +154,8 @@ def TestClusterInit(rapi_user, rapi_secret):
   """gnt-cluster init"""
   master = qa_config.GetMasterNode()
 
-  rapi_dir = os.path.dirname(pathutils.RAPI_USERS_FILE)
+  rapi_users_path = qa_utils.MakeNodePath(master, pathutils.RAPI_USERS_FILE)
+  rapi_dir = os.path.dirname(rapi_users_path)
 
   # First create the RAPI credentials
   fh = tempfile.NamedTemporaryFile()
@@ -164,7 +166,7 @@ def TestClusterInit(rapi_user, rapi_secret):
     tmpru = qa_utils.UploadFile(master.primary, fh.name)
     try:
       AssertCommand(["mkdir", "-p", rapi_dir])
-      AssertCommand(["mv", tmpru, pathutils.RAPI_USERS_FILE])
+      AssertCommand(["mv", tmpru, rapi_users_path])
     finally:
       AssertCommand(["rm", "-f", tmpru])
   finally:
@@ -175,6 +177,8 @@ def TestClusterInit(rapi_user, rapi_secret):
     "gnt-cluster", "init",
     "--primary-ip-version=%d" % qa_config.get("primary_ip_version", 4),
     "--enabled-hypervisors=%s" % ",".join(qa_config.GetEnabledHypervisors()),
+    "--enabled-disk-templates=%s" %
+      ",".join(qa_config.GetEnabledDiskTemplates())
     ]
 
   for spec_type in ("mem-size", "disk-size", "disk-count", "cpu-count",
@@ -182,7 +186,7 @@ def TestClusterInit(rapi_user, rapi_secret):
     for spec_val in ("min", "max", "std"):
       spec = qa_config.get("ispec_%s_%s" %
                            (spec_type.replace("-", "_"), spec_val), None)
-      if spec:
+      if spec is not None:
         cmd.append("--specs-%s=%s=%d" % (spec_type, spec_val, spec))
 
   if master.secondary:
@@ -209,7 +213,12 @@ def TestClusterInit(rapi_user, rapi_secret):
     e_s = False
   qa_config.SetExclusiveStorage(e_s)
 
+  extra_args = qa_config.get("cluster-init-args")
+  if extra_args:
+    cmd.extend(extra_args)
+
   cmd.append(qa_config.get("name"))
+
   AssertCommand(cmd)
 
   cmd = ["gnt-cluster", "modify"]
@@ -382,6 +391,105 @@ def TestClusterModifyDisk():
     AssertCommand(["gnt-cluster", "modify", "-D", param], fail=True)
 
 
+def TestClusterModifyDiskTemplates():
+  """gnt-cluster modify --enabled-disk-templates=..."""
+  enabled_disk_templates = qa_config.GetEnabledDiskTemplates()
+  default_disk_template = qa_config.GetDefaultDiskTemplate()
+
+  _TestClusterModifyDiskTemplatesArguments(default_disk_template,
+                                           enabled_disk_templates)
+
+  _RestoreEnabledDiskTemplates()
+  nodes = qa_config.AcquireManyNodes(2)
+
+  instance_template = enabled_disk_templates[0]
+  instance = qa_instance.CreateInstanceByDiskTemplate(nodes, instance_template)
+
+  _TestClusterModifyUnusedDiskTemplate(instance_template)
+  _TestClusterModifyUsedDiskTemplate(instance_template,
+                                     enabled_disk_templates)
+
+  qa_instance.TestInstanceRemove(instance)
+  _RestoreEnabledDiskTemplates()
+
+
+def _RestoreEnabledDiskTemplates():
+  """Sets the list of enabled disk templates back to the list of enabled disk
+     templates from the QA configuration. This can be used to make sure that
+     the tests that modify the list of disk templates do not interfere with
+     other tests.
+
+  """
+  AssertCommand(
+    ["gnt-cluster", "modify",
+     "--enabled-disk-template=%s" %
+       ",".join(qa_config.GetEnabledDiskTemplates())],
+    fail=False)
+
+
+def _TestClusterModifyDiskTemplatesArguments(default_disk_template,
+                                             enabled_disk_templates):
+  """Tests argument handling of 'gnt-cluster modify' with respect to
+     the parameter '--enabled-disk-templates'. This test is independent
+     of instances.
+
+  """
+  AssertCommand(
+    ["gnt-cluster", "modify",
+     "--enabled-disk-template=%s" %
+       ",".join(enabled_disk_templates)],
+    fail=False)
+
+  # bogus templates
+  AssertCommand(["gnt-cluster", "modify",
+                 "--enabled-disk-templates=pinkbunny"],
+                fail=True)
+
+  # duplicate entries do no harm
+  AssertCommand(
+    ["gnt-cluster", "modify",
+     "--enabled-disk-templates=%s,%s" %
+      (default_disk_template, default_disk_template)],
+    fail=False)
+
+
+def _TestClusterModifyUsedDiskTemplate(instance_template,
+                                       enabled_disk_templates):
+  """Tests that disk templates that are currently in use by instances cannot
+     be disabled on the cluster.
+
+  """
+  # If the list of enabled disk templates contains only one template
+  # we need to add some other templates, because the list of enabled disk
+  # templates can only be set to a non-empty list.
+  new_disk_templates = list(set(enabled_disk_templates)
+                              - set([instance_template]))
+  if not new_disk_templates:
+    new_disk_templates = list(set(constants.DISK_TEMPLATES)
+                                - set([instance_template]))
+  AssertCommand(
+    ["gnt-cluster", "modify",
+     "--enabled-disk-templates=%s" %
+       ",".join(new_disk_templates)],
+    fail=True)
+
+
+def _TestClusterModifyUnusedDiskTemplate(instance_template):
+  """Tests that unused disk templates can be disabled safely."""
+  all_disk_templates = constants.DISK_TEMPLATES
+  AssertCommand(
+    ["gnt-cluster", "modify",
+     "--enabled-disk-templates=%s" %
+       ",".join(all_disk_templates)],
+    fail=False)
+  new_disk_templates = [instance_template]
+  AssertCommand(
+    ["gnt-cluster", "modify",
+     "--enabled-disk-templates=%s" %
+       ",".join(new_disk_templates)],
+    fail=False)
+
+
 def TestClusterModifyBe():
   """gnt-cluster modify -B"""
   for fail, cmd in [
@@ -418,6 +526,171 @@ def TestClusterModifyBe():
     AssertCommand(["gnt-cluster", "modify", "-B", bep])
 
 
+def _GetClusterIPolicy():
+  """Return the run-time values of the cluster-level instance policy.
+
+  @rtype: tuple
+  @return: (policy, specs), where:
+      - policy is a dictionary of the policy values, instance specs excluded
+      - specs is dict of dict, specs[par][key] is a spec value, where key is
+        "min", "max", or "std"
+
+  """
+  info = qa_utils.GetObjectInfo(["gnt-cluster", "info"])
+  policy = info["Instance policy - limits for instances"]
+  ret_specs = {}
+  ret_policy = {}
+  ispec_keys = constants.ISPECS_MINMAX_KEYS | frozenset([constants.ISPECS_STD])
+  for (key, val) in policy.items():
+    if key in ispec_keys:
+      for (par, pval) in val.items():
+        if par == "memory-size":
+          par = "mem-size"
+        d = ret_specs.setdefault(par, {})
+        d[key] = pval
+    else:
+      ret_policy[key] = val
+
+  # Sanity checks
+  assert len(ret_specs) > 0
+  good = all("min" in d and "std" in d and "max" in d
+             for d in ret_specs.values())
+  assert good, "Missing item in specs: %s" % ret_specs
+  assert len(ret_policy) > 0
+  return (ret_policy, ret_specs)
+
+
+def TestClusterModifyIPolicy():
+  """gnt-cluster modify --ipolicy-*"""
+  basecmd = ["gnt-cluster", "modify"]
+  (old_policy, old_specs) = _GetClusterIPolicy()
+  for par in ["vcpu-ratio", "spindle-ratio"]:
+    curr_val = float(old_policy[par])
+    test_values = [
+      (True, 1.0),
+      (True, 1.5),
+      (True, 2),
+      (False, "a"),
+      # Restore the old value
+      (True, curr_val),
+      ]
+    for (good, val) in test_values:
+      cmd = basecmd + ["--ipolicy-%s=%s" % (par, val)]
+      AssertCommand(cmd, fail=not good)
+      if good:
+        curr_val = val
+      # Check the affected parameter
+      (eff_policy, eff_specs) = _GetClusterIPolicy()
+      AssertEqual(float(eff_policy[par]), curr_val)
+      # Check everything else
+      AssertEqual(eff_specs, old_specs)
+      for p in eff_policy.keys():
+        if p == par:
+          continue
+        AssertEqual(eff_policy[p], old_policy[p])
+
+  # Disk templates are treated slightly differently
+  par = "disk-templates"
+  disp_str = "enabled disk templates"
+  curr_val = old_policy[disp_str]
+  test_values = [
+    (True, constants.DT_PLAIN),
+    (True, "%s,%s" % (constants.DT_PLAIN, constants.DT_DRBD8)),
+    (False, "thisisnotadisktemplate"),
+    (False, ""),
+    # Restore the old value
+    (True, curr_val.replace(" ", "")),
+    ]
+  for (good, val) in test_values:
+    cmd = basecmd + ["--ipolicy-%s=%s" % (par, val)]
+    AssertCommand(cmd, fail=not good)
+    if good:
+      curr_val = val
+    # Check the affected parameter
+    (eff_policy, eff_specs) = _GetClusterIPolicy()
+    AssertEqual(eff_policy[disp_str].replace(" ", ""), curr_val)
+    # Check everything else
+    AssertEqual(eff_specs, old_specs)
+    for p in eff_policy.keys():
+      if p == disp_str:
+        continue
+      AssertEqual(eff_policy[p], old_policy[p])
+
+
+def TestClusterSetISpecs(new_specs, fail=False, old_values=None):
+  """Change instance specs.
+
+  @type new_specs: dict of dict
+  @param new_specs: new_specs[par][key], where key is "min", "max", "std". It
+      can be an empty dictionary.
+  @type fail: bool
+  @param fail: if the change is expected to fail
+  @type old_values: tuple
+  @param old_values: (old_policy, old_specs), as returned by
+     L{_GetClusterIPolicy}
+  @return: same as L{_GetClusterIPolicy}
+
+  """
+  if old_values:
+    (old_policy, old_specs) = old_values
+  else:
+    (old_policy, old_specs) = _GetClusterIPolicy()
+  if new_specs:
+    cmd = ["gnt-cluster", "modify"]
+    for (par, keyvals) in new_specs.items():
+      if par == "spindle-use":
+        # ignore spindle-use, which is not settable
+        continue
+      cmd += [
+        "--specs-%s" % par,
+        ",".join(["%s=%s" % (k, v) for (k, v) in keyvals.items()]),
+        ]
+    AssertCommand(cmd, fail=fail)
+  # Check the new state
+  (eff_policy, eff_specs) = _GetClusterIPolicy()
+  AssertEqual(eff_policy, old_policy)
+  if fail:
+    AssertEqual(eff_specs, old_specs)
+  else:
+    for par in eff_specs:
+      for key in eff_specs[par]:
+        if par in new_specs and key in new_specs[par]:
+          AssertEqual(int(eff_specs[par][key]), int(new_specs[par][key]))
+        else:
+          AssertEqual(int(eff_specs[par][key]), int(old_specs[par][key]))
+  return (eff_policy, eff_specs)
+
+
+def TestClusterModifyISpecs():
+  """gnt-cluster modify --specs-*"""
+  params = ["mem-size", "disk-size", "disk-count", "cpu-count", "nic-count"]
+  (cur_policy, cur_specs) = _GetClusterIPolicy()
+  for par in params:
+    test_values = [
+      (True, 0, 4, 12),
+      (True, 4, 4, 12),
+      (True, 4, 12, 12),
+      (True, 4, 4, 4),
+      (False, 4, 0, 12),
+      (False, 4, 16, 12),
+      (False, 4, 4, 0),
+      (False, 12, 4, 4),
+      (False, 12, 4, 0),
+      (False, "a", 4, 12),
+      (False, 0, "a", 12),
+      (False, 0, 4, "a"),
+      # This is to restore the old values
+      (True,
+       cur_specs[par]["min"], cur_specs[par]["std"], cur_specs[par]["max"])
+      ]
+    for (good, mn, st, mx) in test_values:
+      new_vals = {par: {"min": str(mn), "std": str(st), "max": str(mx)}}
+      cur_state = (cur_policy, cur_specs)
+      # We update cur_specs, as we've copied the values to restore already
+      (cur_policy, cur_specs) = TestClusterSetISpecs(new_vals, fail=not good,
+                                                     old_values=cur_state)
+
+
 def TestClusterInfo():
   """gnt-cluster info"""
   AssertCommand(["gnt-cluster", "info"])
@@ -505,7 +778,7 @@ def TestClusterBurnin():
   master = qa_config.GetMasterNode()
 
   options = qa_config.get("options", {})
-  disk_template = options.get("burnin-disk-template", "drbd")
+  disk_template = options.get("burnin-disk-template", constants.DT_DRBD8)
   parallel = options.get("burnin-in-parallel", False)
   check_inst = options.get("burnin-check-instances", False)
   do_rename = options.get("burnin-rename", "")
@@ -527,13 +800,14 @@ def TestClusterBurnin():
 
     script = qa_utils.UploadFile(master.primary, "../tools/burnin")
     try:
+      disks = qa_config.GetDiskOptions()
       # Run burnin
       cmd = [script,
              "--os=%s" % qa_config.get("os"),
              "--minmem-size=%s" % qa_config.get(constants.BE_MINMEM),
              "--maxmem-size=%s" % qa_config.get(constants.BE_MAXMEM),
-             "--disk-size=%s" % ",".join(qa_config.get("disk")),
-             "--disk-growth=%s" % ",".join(qa_config.get("disk-growth")),
+             "--disk-size=%s" % ",".join([d.get("size") for d in disks]),
+             "--disk-growth=%s" % ",".join([d.get("growth") for d in disks]),
              "--disk-template=%s" % disk_template]
       if parallel:
         cmd.append("--parallel")
@@ -570,34 +844,48 @@ def TestClusterMasterFailover():
     failovermaster.Release()
 
 
+def _NodeQueueDrainFile(node):
+  """Returns path to queue drain file for a node.
+
+  """
+  return qa_utils.MakeNodePath(node, pathutils.JOB_QUEUE_DRAIN_FILE)
+
+
+def _AssertDrainFile(node, **kwargs):
+  """Checks for the queue drain file.
+
+  """
+  AssertCommand(["test", "-f", _NodeQueueDrainFile(node)], node=node, **kwargs)
+
+
 def TestClusterMasterFailoverWithDrainedQueue():
   """gnt-cluster master-failover with drained queue"""
-  drain_check = ["test", "-f", pathutils.JOB_QUEUE_DRAIN_FILE]
-
   master = qa_config.GetMasterNode()
   failovermaster = qa_config.AcquireNode(exclude=master)
 
   # Ensure queue is not drained
   for node in [master, failovermaster]:
-    AssertCommand(drain_check, node=node, fail=True)
+    _AssertDrainFile(node, fail=True)
 
   # Drain queue on failover master
-  AssertCommand(["touch", pathutils.JOB_QUEUE_DRAIN_FILE], node=failovermaster)
+  AssertCommand(["touch", _NodeQueueDrainFile(failovermaster)],
+                node=failovermaster)
 
   cmd = ["gnt-cluster", "master-failover"]
   try:
-    AssertCommand(drain_check, node=failovermaster)
+    _AssertDrainFile(failovermaster)
     AssertCommand(cmd, node=failovermaster)
-    AssertCommand(drain_check, fail=True)
-    AssertCommand(drain_check, node=failovermaster, fail=True)
+    _AssertDrainFile(master, fail=True)
+    _AssertDrainFile(failovermaster, fail=True)
 
     # Back to original master node
     AssertCommand(cmd, node=master)
   finally:
     failovermaster.Release()
 
-  AssertCommand(drain_check, fail=True)
-  AssertCommand(drain_check, node=failovermaster, fail=True)
+  # Ensure queue is not drained
+  for node in [master, failovermaster]:
+    _AssertDrainFile(node, fail=True)
 
 
 def TestClusterCopyfile():
@@ -656,10 +944,11 @@ def TestSetExclStorCluster(newvalue):
   @return: The old value of exclusive_storage
 
   """
-  oldvalue = _GetBoolClusterField("exclusive_storage")
+  es_path = ["Default node parameters", "exclusive_storage"]
+  oldvalue = _GetClusterField(es_path)
   AssertCommand(["gnt-cluster", "modify", "--node-parameters",
                  "exclusive_storage=%s" % newvalue])
-  effvalue = _GetBoolClusterField("exclusive_storage")
+  effvalue = _GetClusterField(es_path)
   if effvalue != newvalue:
     raise qa_error.Error("exclusive_storage has the wrong value: %s instead"
                          " of %s" % (effvalue, newvalue))
@@ -667,24 +956,6 @@ def TestSetExclStorCluster(newvalue):
   return oldvalue
 
 
-def _BuildSetESCmd(value, node_name):
-  return ["gnt-node", "modify", "--node-parameters",
-          "exclusive_storage=%s" % value, node_name]
-
-
-def TestExclStorSingleNode(node):
-  """cluster-verify reports exclusive_storage set only on one node.
-
-  """
-  node_name = node.primary
-  es_val = _GetBoolClusterField("exclusive_storage")
-  assert not es_val
-  AssertCommand(_BuildSetESCmd(True, node_name))
-  AssertClusterVerify(fail=True, errors=[constants.CV_EGROUPMIXEDESFLAG])
-  AssertCommand(_BuildSetESCmd("default", node_name))
-  AssertClusterVerify()
-
-
 def TestExclStorSharedPv(node):
   """cluster-verify reports LVs that share the same PV with exclusive_storage.