Add support for vm_capable in cluster verify

The method to make vm_capable integrate easily into cluster verify is as follows:

- we add a new NV_VMNODES that represents *non*-vm-capable nodes
- the LU populates this list (it's expected that non-vm_capable nodes
  are few compared to vm_capable nodes)
- backend skips the checks that are related to VM hosting
- in the LU, we reorder the VM-related checks so that they occur after
  the non-VM (generic) tests, and we only execute them conditionally

Additionally, we add some support to the instance checks to detect
instances living on bad nodes.

Signed-off-by: Iustin Pop <iustin@google.com>
Reviewed-by: Michael Hanselmann <hansmi@google.com>

diff --git a/lib/backend.py b/lib/backend.py
index 3df6396..4bf94fb 100644 (file)
--- a/lib/backend.py
+++ b/lib/backend.py
@@ -491,8 +491,9 @@ def VerifyNode(what, cluster_name):
   result = {}
   my_name = netutils.Hostname.GetSysName()
   port = netutils.GetDaemonPort(constants.NODED)
+  vm_capable = my_name not in what.get(constants.NV_VMNODES, [])
 
-  if constants.NV_HYPERVISOR in what:
+  if constants.NV_HYPERVISOR in what and vm_capable:
     result[constants.NV_HYPERVISOR] = tmp = {}
     for hv_name in what[constants.NV_HYPERVISOR]:
       try:
@@ -547,14 +548,14 @@ def VerifyNode(what, cluster_name):
     result[constants.NV_MASTERIP] = netutils.TcpPing(master_ip, port,
                                                   source=source)
 
-  if constants.NV_LVLIST in what:
+  if constants.NV_LVLIST in what and vm_capable:
     try:
       val = GetVolumeList(what[constants.NV_LVLIST])
     except RPCFail, err:
       val = str(err)
     result[constants.NV_LVLIST] = val
 
-  if constants.NV_INSTANCELIST in what:
+  if constants.NV_INSTANCELIST in what and vm_capable:
     # GetInstanceList can fail
     try:
       val = GetInstanceList(what[constants.NV_INSTANCELIST])
@@ -562,10 +563,10 @@ def VerifyNode(what, cluster_name):
       val = str(err)
     result[constants.NV_INSTANCELIST] = val
 
-  if constants.NV_VGLIST in what:
+  if constants.NV_VGLIST in what and vm_capable:
     result[constants.NV_VGLIST] = utils.ListVolumeGroups()
 
-  if constants.NV_PVLIST in what:
+  if constants.NV_PVLIST in what and vm_capable:
     result[constants.NV_PVLIST] = \
       bdev.LogicalVolume.GetPVInfo(what[constants.NV_PVLIST],
                                    filter_allocatable=False)
@@ -574,11 +575,11 @@ def VerifyNode(what, cluster_name):
     result[constants.NV_VERSION] = (constants.PROTOCOL_VERSION,
                                     constants.RELEASE_VERSION)
 
-  if constants.NV_HVINFO in what:
+  if constants.NV_HVINFO in what and vm_capable:
     hyper = hypervisor.GetHypervisor(what[constants.NV_HVINFO])
     result[constants.NV_HVINFO] = hyper.GetNodeInfo()
 
-  if constants.NV_DRBDLIST in what:
+  if constants.NV_DRBDLIST in what and vm_capable:
     try:
       used_minors = bdev.DRBD8.GetUsedDevs().keys()
     except errors.BlockDeviceError, err:
@@ -586,7 +587,7 @@ def VerifyNode(what, cluster_name):
       used_minors = str(err)
     result[constants.NV_DRBDLIST] = used_minors
 
-  if constants.NV_DRBDHELPER in what:
+  if constants.NV_DRBDHELPER in what and vm_capable:
     status = True
     try:
       payload = bdev.BaseDRBD.GetUsermodeHelper()
@@ -611,7 +612,7 @@ def VerifyNode(what, cluster_name):
   if constants.NV_TIME in what:
     result[constants.NV_TIME] = utils.SplitTime(time.time())
 
-  if constants.NV_OSLIST in what:
+  if constants.NV_OSLIST in what and vm_capable:
     result[constants.NV_OSLIST] = DiagnoseOS()
 
   return result

diff --git a/lib/cmdlib.py b/lib/cmdlib.py
index 6152f36..31dc66f 100644 (file)
--- a/lib/cmdlib.py
+++ b/lib/cmdlib.py
@@ -1219,9 +1219,11 @@ class LUVerifyCluster(LogicalUnit):
     @ivar os_fail: whether the RPC call didn't return valid OS data
     @type oslist: list
     @ivar oslist: list of OSes as diagnosed by DiagnoseOS
+    @type vm_capable: boolean
+    @ivar vm_capable: whether the node can host instances
 
     """
-    def __init__(self, offline=False, name=None):
+    def __init__(self, offline=False, name=None, vm_capable=True):
       self.name = name
       self.volumes = {}
       self.instances = []
@@ -1231,6 +1233,7 @@ class LUVerifyCluster(LogicalUnit):
       self.mfree = 0
       self.dfree = 0
       self.offline = offline
+      self.vm_capable = vm_capable
       self.rpc_fail = False
       self.lvm_fail = False
       self.hyp_fail = False
@@ -1335,7 +1338,7 @@ class LUVerifyCluster(LogicalUnit):
                   code=self.ETYPE_WARNING)
 
     hyp_result = nresult.get(constants.NV_HYPERVISOR, None)
-    if isinstance(hyp_result, dict):
+    if ninfo.vm_capable and isinstance(hyp_result, dict):
       for hv_name, hv_result in hyp_result.iteritems():
         test = hv_result is not None
         _ErrorIf(test, self.ENODEHV, node,
@@ -2010,6 +2013,7 @@ class LUVerifyCluster(LogicalUnit):
       constants.NV_TIME: None,
       constants.NV_MASTERIP: (master_node, master_ip),
       constants.NV_OSLIST: None,
+      constants.NV_VMNODES: self.cfg.GetNonVmCapableNodeList(),
       }
 
     if vg_name is not None:
@@ -2023,7 +2027,8 @@ class LUVerifyCluster(LogicalUnit):
 
     # Build our expected cluster state
     node_image = dict((node.name, self.NodeImage(offline=node.offline,
-                                                 name=node.name))
+                                                 name=node.name,
+                                                 vm_capable=node.vm_capable))
                       for node in nodeinfo)
 
     for instance in instancelist:
@@ -2100,22 +2105,24 @@ class LUVerifyCluster(LogicalUnit):
       nresult = all_nvinfo[node].payload
 
       nimg.call_ok = self._VerifyNode(node_i, nresult)
+      self._VerifyNodeTime(node_i, nresult, nvinfo_starttime, nvinfo_endtime)
       self._VerifyNodeNetwork(node_i, nresult)
-      self._VerifyNodeLVM(node_i, nresult, vg_name)
       self._VerifyNodeFiles(node_i, nresult, file_names, local_checksums,
                             master_files)
-      self._VerifyNodeDrbd(node_i, nresult, instanceinfo, drbd_helper,
-                           all_drbd_map)
-      self._VerifyNodeTime(node_i, nresult, nvinfo_starttime, nvinfo_endtime)
 
-      self._UpdateNodeVolumes(node_i, nresult, nimg, vg_name)
-      self._UpdateNodeInstances(node_i, nresult, nimg)
-      self._UpdateNodeInfo(node_i, nresult, nimg, vg_name)
-      self._UpdateNodeOS(node_i, nresult, nimg)
-      if not nimg.os_fail:
-        if refos_img is None:
-          refos_img = nimg
-        self._VerifyNodeOS(node_i, nimg, refos_img)
+      if nimg.vm_capable:
+        self._VerifyNodeLVM(node_i, nresult, vg_name)
+        self._VerifyNodeDrbd(node_i, nresult, instanceinfo, drbd_helper,
+                             all_drbd_map)
+
+        self._UpdateNodeVolumes(node_i, nresult, nimg, vg_name)
+        self._UpdateNodeInstances(node_i, nresult, nimg)
+        self._UpdateNodeInfo(node_i, nresult, nimg, vg_name)
+        self._UpdateNodeOS(node_i, nresult, nimg)
+        if not nimg.os_fail:
+          if refos_img is None:
+            refos_img = nimg
+          self._VerifyNodeOS(node_i, nimg, refos_img)
 
     feedback_fn("* Verifying instance status")
     for instance in instancelist:
@@ -2162,10 +2169,12 @@ class LUVerifyCluster(LogicalUnit):
       _ErrorIf(inst_nodes_offline, self.EINSTANCEBADNODE, instance,
                "instance lives on offline node(s) %s",
                utils.CommaJoin(inst_nodes_offline))
-      # ... or ghost nodes
+      # ... or ghost/non-vm_capable nodes
       for node in inst_config.all_nodes:
         _ErrorIf(node_image[node].ghost, self.EINSTANCEBADNODE, instance,
                  "instance lives on ghost node %s", node)
+        _ErrorIf(not node_image[node].vm_capable, self.EINSTANCEBADNODE,
+                 instance, "instance lives on non-vm_capable node %s", node)
 
     feedback_fn("* Verifying orphan volumes")
     reserved = utils.FieldSet(*cluster.reserved_lvs)

diff --git a/lib/constants.py b/lib/constants.py
index c581bea..71abd6d 100644 (file)
--- a/lib/constants.py
+++ b/lib/constants.py
@@ -785,6 +785,7 @@ NV_PVLIST = "pvlist"
 NV_TIME = "time"
 NV_VERSION = "version"
 NV_VGLIST = "vglist"
+NV_VMNODES = "vmnodes"
 
 # SSL certificate check constants (in days)
 SSL_CERT_EXPIRATION_WARN = 30