Implement BuildHooksEnv for NoHooksLU

[ganeti-local] / lib / config.py

diff --git a/lib/config.py b/lib/config.py
index 1d792d1..712f599 100644 (file)
--- a/lib/config.py
+++ b/lib/config.py
@@ -79,6 +79,7 @@ class ConfigWriter:
       self._cfg_file = cfg_file
     self._temporary_ids = set()
     self._temporary_drbds = {}
       self._cfg_file = cfg_file
     self._temporary_ids = set()
     self._temporary_drbds = {}

+    self._temporary_macs = set()

     # Note: in order to prevent errors when resolving our name in
     # _DistributeConfig, we compute it here once and reuse it; it's
     # better to raise an error before starting to modify the config
     # Note: in order to prevent errors when resolving our name in
     # _DistributeConfig, we compute it here once and reuse it; it's
     # better to raise an error before starting to modify the config

@@ -110,11 +111,12 @@ class ConfigWriter:
       byte2 = random.randrange(0, 256)
       byte3 = random.randrange(0, 256)
       mac = "%s:%02x:%02x:%02x" % (prefix, byte1, byte2, byte3)
       byte2 = random.randrange(0, 256)
       byte3 = random.randrange(0, 256)
       mac = "%s:%02x:%02x:%02x" % (prefix, byte1, byte2, byte3)

-      if mac not in all_macs:
+      if mac not in all_macs and mac not in self._temporary_macs:

         break
       retries -= 1
     else:
       raise errors.ConfigurationError("Can't generate unique MAC")
         break
       retries -= 1
     else:
       raise errors.ConfigurationError("Can't generate unique MAC")

+    self._temporary_macs.add(mac)

     return mac
 
   @locking.ssynchronized(_config_lock, shared=1)
     return mac
 
   @locking.ssynchronized(_config_lock, shared=1)

@@ -126,7 +128,7 @@ class ConfigWriter:
 
     """
     all_macs = self._AllMACs()
 
     """
     all_macs = self._AllMACs()

-    return mac in all_macs
+    return mac in all_macs or mac in self._temporary_macs

 
   @locking.ssynchronized(_config_lock, shared=1)
   def GenerateDRBDSecret(self):
 
   @locking.ssynchronized(_config_lock, shared=1)
   def GenerateDRBDSecret(self):

@@ -146,7 +148,7 @@ class ConfigWriter:
       raise errors.ConfigurationError("Can't generate unique DRBD secret")
     return secret
 
       raise errors.ConfigurationError("Can't generate unique DRBD secret")
     return secret
 

-  def _ComputeAllLVs(self):
+  def _AllLVs(self):

     """Compute the list of all LVs.
 
     """
     """Compute the list of all LVs.
 
     """

@@ -157,6 +159,23 @@ class ConfigWriter:
         lvnames.update(lv_list)
     return lvnames
 
         lvnames.update(lv_list)
     return lvnames
 

+  def _AllIDs(self, include_temporary):
+    """Compute the list of all UUIDs and names we have.
+
+    @type include_temporary: boolean
+    @param include_temporary: whether to include the _temporary_ids set
+    @rtype: set
+    @return: a set of IDs
+
+    """
+    existing = set()
+    if include_temporary:
+      existing.update(self._temporary_ids)
+    existing.update(self._AllLVs())
+    existing.update(self._config_data.instances.keys())
+    existing.update(self._config_data.nodes.keys())
+    return existing
+

   @locking.ssynchronized(_config_lock, shared=1)
   def GenerateUniqueID(self, exceptions=None):
     """Generate an unique disk name.
   @locking.ssynchronized(_config_lock, shared=1)
   def GenerateUniqueID(self, exceptions=None):
     """Generate an unique disk name.

@@ -173,11 +192,7 @@ class ConfigWriter:
     @return: the unique id
 
     """
     @return: the unique id
 
     """

-    existing = set()
-    existing.update(self._temporary_ids)
-    existing.update(self._ComputeAllLVs())
-    existing.update(self._config_data.instances.keys())
-    existing.update(self._config_data.nodes.keys())
+    existing = self._AllIDs(include_temporary=True)

     if exceptions is not None:
       existing.update(exceptions)
     retries = 64
     if exceptions is not None:
       existing.update(exceptions)
     retries = 64

@@ -191,6 +206,13 @@ class ConfigWriter:
     self._temporary_ids.add(unique_id)
     return unique_id
 
     self._temporary_ids.add(unique_id)
     return unique_id
 

+  def _CleanupTemporaryIDs(self):
+    """Cleanups the _temporary_ids structure.
+
+    """
+    existing = self._AllIDs(include_temporary=False)
+    self._temporary_ids = self._temporary_ids - existing
+

   def _AllMACs(self):
     """Return all MACs present in the config.
 
   def _AllMACs(self):
     """Return all MACs present in the config.
 

@@ -227,6 +249,36 @@ class ConfigWriter:
 
     return result
 
 
     return result
 

+  def _CheckDiskIDs(self, disk, l_ids, p_ids):
+    """Compute duplicate disk IDs
+
+    @type disk: L{objects.Disk}
+    @param disk: the disk at which to start searching
+    @type l_ids: list
+    @param l_ids: list of current logical ids
+    @type p_ids: list
+    @param p_ids: list of current physical ids
+    @rtype: list
+    @return: a list of error messages
+
+    """
+    result = []
+    if disk.logical_id is not None:
+      if disk.logical_id in l_ids:
+        result.append("duplicate logical id %s" % str(disk.logical_id))
+      else:
+        l_ids.append(disk.logical_id)
+    if disk.physical_id is not None:
+      if disk.physical_id in p_ids:
+        result.append("duplicate physical id %s" % str(disk.physical_id))
+      else:
+        p_ids.append(disk.physical_id)
+
+    if disk.children:
+      for child in disk.children:
+        result.extend(self._CheckDiskIDs(child, l_ids, p_ids))
+    return result
+

   def _UnlockedVerifyConfig(self):
     """Verify function.
 
   def _UnlockedVerifyConfig(self):
     """Verify function.
 

@@ -239,6 +291,22 @@ class ConfigWriter:
     seen_macs = []
     ports = {}
     data = self._config_data
     seen_macs = []
     ports = {}
     data = self._config_data

+    seen_lids = []
+    seen_pids = []
+
+    # global cluster checks
+    if not data.cluster.enabled_hypervisors:
+      result.append("enabled hypervisors list doesn't have any entries")
+    invalid_hvs = set(data.cluster.enabled_hypervisors) - constants.HYPER_TYPES
+    if invalid_hvs:
+      result.append("enabled hypervisors contains invalid entries: %s" %
+                    invalid_hvs)
+
+    if data.cluster.master_node not in data.nodes:
+      result.append("cluster has invalid primary node '%s'" %
+                    data.cluster.master_node)
+
+    # per-instance checks

     for instance_name in data.instances:
       instance = data.instances[instance_name]
       if instance.primary_node not in data.nodes:
     for instance_name in data.instances:
       instance = data.instances[instance_name]
       if instance.primary_node not in data.nodes:

@@ -269,6 +337,12 @@ class ConfigWriter:
           ports[net_port] = []
         ports[net_port].append((instance.name, "network port"))
 
           ports[net_port] = []
         ports[net_port].append((instance.name, "network port"))
 

+      # instance disk verify
+      for idx, disk in enumerate(instance.disks):
+        result.extend(["instance '%s' disk %d error: %s" %
+                       (instance.name, idx, msg) for msg in disk.Verify()])
+        result.extend(self._CheckDiskIDs(disk, seen_lids, seen_pids))
+

     # cluster-wide pool of free ports
     for free_port in data.cluster.tcpudp_port_pool:
       if free_port not in ports:
     # cluster-wide pool of free ports
     for free_port in data.cluster.tcpudp_port_pool:
       if free_port not in ports:

@@ -299,6 +373,14 @@ class ConfigWriter:
       result.append("Not enough master candidates: actual %d, target %d" %
                     (mc_now, mc_max))
 
       result.append("Not enough master candidates: actual %d, target %d" %
                     (mc_now, mc_max))
 

+    # node checks
+    for node in data.nodes.values():
+      if [node.master_candidate, node.drained, node.offline].count(True) > 1:
+        result.append("Node %s state is invalid: master_candidate=%s,"
+                      " drain=%s, offline=%s" %
+                      (node.name, node.master_candidate, node.drain,
+                       node.offline))
+

     # drbd minors check
     d_map, duplicates = self._UnlockedComputeDRBDMap()
     for node, minor, instance_a, instance_b in duplicates:
     # drbd minors check
     d_map, duplicates = self._UnlockedComputeDRBDMap()
     for node, minor, instance_a, instance_b in duplicates:

@@ -426,8 +508,8 @@ class ConfigWriter:
     def _AppendUsedPorts(instance_name, disk, used):
       duplicates = []
       if disk.dev_type == constants.LD_DRBD8 and len(disk.logical_id) >= 5:
     def _AppendUsedPorts(instance_name, disk, used):
       duplicates = []
       if disk.dev_type == constants.LD_DRBD8 and len(disk.logical_id) >= 5:

-        nodeA, nodeB, dummy, minorA, minorB = disk.logical_id[:5]
-        for node, port in ((nodeA, minorA), (nodeB, minorB)):
+        node_a, node_b, _, minor_a, minor_b = disk.logical_id[:5]
+        for node, port in ((node_a, minor_a), (node_b, minor_b)):

           assert node in used, ("Node '%s' of instance '%s' not found"
                                 " in node list" % (node, instance_name))
           if port in used[node]:
           assert node in used, ("Node '%s' of instance '%s' not found"
                                 " in node list" % (node, instance_name))
           if port in used[node]:

@@ -639,9 +721,18 @@ class ConfigWriter:
       all_lvs = instance.MapLVsByNode()
       logging.info("Instance '%s' DISK_LAYOUT: %s", instance.name, all_lvs)
 
       all_lvs = instance.MapLVsByNode()
       logging.info("Instance '%s' DISK_LAYOUT: %s", instance.name, all_lvs)
 

+    all_macs = self._AllMACs()
+    for nic in instance.nics:
+      if nic.mac in all_macs:
+        raise errors.ConfigurationError("Cannot add instance %s:"
+          " MAC address '%s' already in use." % (instance.name, nic.mac))
+

     instance.serial_no = 1
     self._config_data.instances[instance.name] = instance
     instance.serial_no = 1
     self._config_data.instances[instance.name] = instance

+    self._config_data.cluster.serial_no += 1

     self._UnlockedReleaseDRBDMinors(instance.name)
     self._UnlockedReleaseDRBDMinors(instance.name)

+    for nic in instance.nics:
+      self._temporary_macs.discard(nic.mac)

     self._WriteConfig()
 
   def _SetInstanceStatus(self, instance_name, status):
     self._WriteConfig()
 
   def _SetInstanceStatus(self, instance_name, status):

@@ -675,6 +766,7 @@ class ConfigWriter:
     if instance_name not in self._config_data.instances:
       raise errors.ConfigurationError("Unknown instance '%s'" % instance_name)
     del self._config_data.instances[instance_name]
     if instance_name not in self._config_data.instances:
       raise errors.ConfigurationError("Unknown instance '%s'" % instance_name)
     del self._config_data.instances[instance_name]

+    self._config_data.cluster.serial_no += 1

     self._WriteConfig()
 
   @locking.ssynchronized(_config_lock)
     self._WriteConfig()
 
   @locking.ssynchronized(_config_lock)

@@ -738,7 +830,7 @@ class ConfigWriter:
                                     self._config_data.instances.keys())
 
   def _UnlockedGetInstanceInfo(self, instance_name):
                                     self._config_data.instances.keys())
 
   def _UnlockedGetInstanceInfo(self, instance_name):

-    """Returns informations about an instance.
+    """Returns information about an instance.

 
     This function is for internal use, when the config lock is already held.
 
 
     This function is for internal use, when the config lock is already held.
 

@@ -750,9 +842,9 @@ class ConfigWriter:
 
   @locking.ssynchronized(_config_lock, shared=1)
   def GetInstanceInfo(self, instance_name):
 
   @locking.ssynchronized(_config_lock, shared=1)
   def GetInstanceInfo(self, instance_name):

-    """Returns informations about an instance.
+    """Returns information about an instance.

 
 

-    It takes the information from the configuration file. Other informations of
+    It takes the information from the configuration file. Other information of

     an instance are taken from the live systems.
 
     @param instance_name: name of the instance, e.g.
     an instance are taken from the live systems.
 
     @param instance_name: name of the instance, e.g.

@@ -769,7 +861,7 @@ class ConfigWriter:
     """Get the configuration of all instances.
 
     @rtype: dict
     """Get the configuration of all instances.
 
     @rtype: dict

-    @returns: dict of (instance, instance_info), where instance_info is what
+    @return: dict of (instance, instance_info), where instance_info is what

               would GetInstanceInfo return for the node
 
     """
               would GetInstanceInfo return for the node
 
     """

@@ -887,16 +979,20 @@ class ConfigWriter:
                     for node in self._UnlockedGetNodeList()])
     return my_dict
 
                     for node in self._UnlockedGetNodeList()])
     return my_dict
 

-  def _UnlockedGetMasterCandidateStats(self):
+  def _UnlockedGetMasterCandidateStats(self, exceptions=None):

     """Get the number of current and maximum desired and possible candidates.
 
     """Get the number of current and maximum desired and possible candidates.
 

+    @type exceptions: list
+    @param exceptions: if passed, list of nodes that should be ignored

     @rtype: tuple
     @return: tuple of (current, desired and possible)
 
     """
     mc_now = mc_max = 0
     @rtype: tuple
     @return: tuple of (current, desired and possible)
 
     """
     mc_now = mc_max = 0

-    for node in self._config_data.nodes.itervalues():
-      if not node.offline:
+    for node in self._config_data.nodes.values():
+      if exceptions and node.name in exceptions:
+        continue
+      if not (node.offline or node.drained):

         mc_max += 1
       if node.master_candidate:
         mc_now += 1
         mc_max += 1
       if node.master_candidate:
         mc_now += 1

@@ -904,16 +1000,18 @@ class ConfigWriter:
     return (mc_now, mc_max)
 
   @locking.ssynchronized(_config_lock, shared=1)
     return (mc_now, mc_max)
 
   @locking.ssynchronized(_config_lock, shared=1)

-  def GetMasterCandidateStats(self):
+  def GetMasterCandidateStats(self, exceptions=None):

     """Get the number of current and maximum possible candidates.
 
     This is just a wrapper over L{_UnlockedGetMasterCandidateStats}.
 
     """Get the number of current and maximum possible candidates.
 
     This is just a wrapper over L{_UnlockedGetMasterCandidateStats}.
 

+    @type exceptions: list
+    @param exceptions: if passed, list of nodes that should be ignored

     @rtype: tuple
     @return: tuple of (current, max)
 
     """
     @rtype: tuple
     @return: tuple of (current, max)
 
     """

-    return self._UnlockedGetMasterCandidateStats()
+    return self._UnlockedGetMasterCandidateStats(exceptions)

 
   @locking.ssynchronized(_config_lock)
   def MaintainCandidatePool(self):
 
   @locking.ssynchronized(_config_lock)
   def MaintainCandidatePool(self):

@@ -932,7 +1030,7 @@ class ConfigWriter:
         if mc_now >= mc_max:
           break
         node = self._config_data.nodes[name]
         if mc_now >= mc_max:
           break
         node = self._config_data.nodes[name]

-        if node.master_candidate or node.offline:
+        if node.master_candidate or node.offline or node.drained:

           continue
         mod_list.append(node)
         node.master_candidate = True
           continue
         mod_list.append(node)
         node.master_candidate = True

@@ -974,6 +1072,10 @@ class ConfigWriter:
         not hasattr(data.cluster, 'rsahostkeypub')):
       raise errors.ConfigurationError("Incomplete configuration"
                                       " (missing cluster.rsahostkeypub)")
         not hasattr(data.cluster, 'rsahostkeypub')):
       raise errors.ConfigurationError("Incomplete configuration"
                                       " (missing cluster.rsahostkeypub)")

+
+    # Upgrade configuration if needed
+    data.UpgradeConfig()
+

     self._config_data = data
     # reset the last serial as -1 so that the next write will cause
     # ssconf update
     self._config_data = data
     # reset the last serial as -1 so that the next write will cause
     # ssconf update

@@ -1019,6 +1121,10 @@ class ConfigWriter:
     """Write the configuration data to persistent storage.
 
     """
     """Write the configuration data to persistent storage.
 
     """

+    # first, cleanup the _temporary_ids set, if an ID is now in the
+    # other objects it should be discarded to prevent unbounded growth
+    # of that structure
+    self._CleanupTemporaryIDs()

     config_errors = self._UnlockedVerifyConfig()
     if config_errors:
       raise errors.ConfigurationError("Configuration data is not"
     config_errors = self._UnlockedVerifyConfig()
     if config_errors:
       raise errors.ConfigurationError("Configuration data is not"

@@ -1059,16 +1165,21 @@ class ConfigWriter:
 
     """
     fn = "\n".join
 
     """
     fn = "\n".join

+    instance_names = utils.NiceSort(self._UnlockedGetInstanceList())

     node_names = utils.NiceSort(self._UnlockedGetNodeList())
     node_info = [self._UnlockedGetNodeInfo(name) for name in node_names]
 
     node_names = utils.NiceSort(self._UnlockedGetNodeList())
     node_info = [self._UnlockedGetNodeInfo(name) for name in node_names]
 

+    instance_data = fn(instance_names)

     off_data = fn(node.name for node in node_info if node.offline)
     off_data = fn(node.name for node in node_info if node.offline)

+    on_data = fn(node.name for node in node_info if not node.offline)

     mc_data = fn(node.name for node in node_info if node.master_candidate)
     node_data = fn(node_names)
 
     cluster = self._config_data.cluster
     mc_data = fn(node.name for node in node_info if node.master_candidate)
     node_data = fn(node_names)
 
     cluster = self._config_data.cluster

+    cluster_tags = fn(cluster.GetTags())

     return {
       constants.SS_CLUSTER_NAME: cluster.cluster_name,
     return {
       constants.SS_CLUSTER_NAME: cluster.cluster_name,

+      constants.SS_CLUSTER_TAGS: cluster_tags,

       constants.SS_FILE_STORAGE_DIR: cluster.file_storage_dir,
       constants.SS_MASTER_CANDIDATES: mc_data,
       constants.SS_MASTER_IP: cluster.master_ip,
       constants.SS_FILE_STORAGE_DIR: cluster.file_storage_dir,
       constants.SS_MASTER_CANDIDATES: mc_data,
       constants.SS_MASTER_IP: cluster.master_ip,

@@ -1076,35 +1187,11 @@ class ConfigWriter:
       constants.SS_MASTER_NODE: cluster.master_node,
       constants.SS_NODE_LIST: node_data,
       constants.SS_OFFLINE_NODES: off_data,
       constants.SS_MASTER_NODE: cluster.master_node,
       constants.SS_NODE_LIST: node_data,
       constants.SS_OFFLINE_NODES: off_data,

+      constants.SS_ONLINE_NODES: on_data,
+      constants.SS_INSTANCE_LIST: instance_data,

       constants.SS_RELEASE_VERSION: constants.RELEASE_VERSION,
       }
 
       constants.SS_RELEASE_VERSION: constants.RELEASE_VERSION,
       }
 

-  @locking.ssynchronized(_config_lock)
-  def InitConfig(self, version, cluster_config, master_node_config):
-    """Create the initial cluster configuration.
-
-    It will contain the current node, which will also be the master
-    node, and no instances.
-
-    @type version: int
-    @param version: Configuration version
-    @type cluster_config: objects.Cluster
-    @param cluster_config: Cluster configuration
-    @type master_node_config: objects.Node
-    @param master_node_config: Master node configuration
-
-    """
-    nodes = {
-      master_node_config.name: master_node_config,
-      }
-
-    self._config_data = objects.ConfigData(version=version,
-                                           cluster=cluster_config,
-                                           nodes=nodes,
-                                           instances={},
-                                           serial_no=1)
-    self._WriteConfig()
-

   @locking.ssynchronized(_config_lock, shared=1)
   def GetVGName(self):
     """Return the volume group name.
   @locking.ssynchronized(_config_lock, shared=1)
   def GetVGName(self):
     """Return the volume group name.

@@ -1137,7 +1224,7 @@ class ConfigWriter:
 
   @locking.ssynchronized(_config_lock, shared=1)
   def GetClusterInfo(self):
 
   @locking.ssynchronized(_config_lock, shared=1)
   def GetClusterInfo(self):

-    """Returns informations about the cluster
+    """Returns information about the cluster

 
     @rtype: L{objects.Cluster}
     @return: the cluster object
 
     @rtype: L{objects.Cluster}
     @return: the cluster object

@@ -1185,5 +1272,7 @@ class ConfigWriter:
 
     if isinstance(target, objects.Instance):
       self._UnlockedReleaseDRBDMinors(target.name)
 
     if isinstance(target, objects.Instance):
       self._UnlockedReleaseDRBDMinors(target.name)

+      for nic in target.nics:
+        self._temporary_macs.discard(nic.mac)

 
     self._WriteConfig()
 
     self._WriteConfig()