gnt-instance: Adding instance policy to failover

Signed-off-by: René Nussbaumer <rn@google.com>
Reviewed-by: Michael Hanselmann <hansmi@google.com>

diff --git a/lib/cli.py b/lib/cli.py
index 30f6771..00cb0da 100644 (file)
--- a/lib/cli.py
+++ b/lib/cli.py
@@ -192,6 +192,7 @@ __all__ = [
   "YES_DOIT_OPT",
   "DISK_STATE_OPT",
   "HV_STATE_OPT",
+  "IGNORE_IPOLICY_OPT",
   # Generic functions for CLI programs
   "ConfirmOperation",
   "GenericMain",
@@ -1334,6 +1335,10 @@ HV_STATE_OPT = cli_option("--hypervisor-state", default=[], dest="hv_state",
                                 " format hypervisor:option=value,..."),
                           type="identkeyval")
 
+IGNORE_IPOLICY_OPT = cli_option("--ignore-ipolicy", dest="ignore_ipolicy",
+                                action="store_true", default=False,
+                                help="Ignore instance policy violations")
+
 
 #: Options provided by all commands
 COMMON_OPTS = [DEBUG_OPT]

diff --git a/lib/client/gnt_instance.py b/lib/client/gnt_instance.py
index f26063a..044f5c0 100644 (file)
--- a/lib/client/gnt_instance.py
+++ b/lib/client/gnt_instance.py
@@ -791,7 +791,8 @@ def FailoverInstance(opts, args):
                                   ignore_consistency=opts.ignore_consistency,
                                   shutdown_timeout=opts.shutdown_timeout,
                                   iallocator=iallocator,
-                                  target_node=target_node)
+                                  target_node=target_node,
+                                  ignore_ipolicy=opts.ignore_ipolicy)
   SubmitOrSend(op, opts, cl=cl)
   return 0
 
@@ -1436,7 +1437,8 @@ commands = {
   "failover": (
     FailoverInstance, ARGS_ONE_INSTANCE,
     [FORCE_OPT, IGNORE_CONSIST_OPT, SUBMIT_OPT, SHUTDOWN_TIMEOUT_OPT,
-     DRY_RUN_OPT, PRIORITY_OPT, DST_NODE_OPT, IALLOCATOR_OPT],
+     DRY_RUN_OPT, PRIORITY_OPT, DST_NODE_OPT, IALLOCATOR_OPT,
+     IGNORE_IPOLICY_OPT],
     "[-f] <instance>", "Stops the instance, changes its primary node and"
     " (if it was originally running) starts it on the new node"
     " (the secondary for mirrored instances or any node"

diff --git a/lib/cmdlib.py b/lib/cmdlib.py
index 8591e8f..8d7f2ba 100644 (file)
--- a/lib/cmdlib.py
+++ b/lib/cmdlib.py
@@ -7244,7 +7244,8 @@ class LUInstanceFailover(LogicalUnit):
                                        cleanup=False,
                                        failover=True,
                                        ignore_consistency=ignore_consistency,
-                                       shutdown_timeout=shutdown_timeout)
+                                       shutdown_timeout=shutdown_timeout,
+                                       ignore_ipolicy=self.op.ignore_ipolicy)
     self.tasklets = [self._migrater]
 
   def DeclareLocks(self, level):
@@ -7644,6 +7645,8 @@ class TLMigrateInstance(Tasklet):
                             and target node
   @type shutdown_timeout: int
   @ivar shutdown_timeout: In case of failover timeout of the shutdown
+  @type ignore_ipolicy: bool
+  @ivar ignore_ipolicy: If true, we can ignore instance policy when migrating
 
   """
 
@@ -7654,7 +7657,8 @@ class TLMigrateInstance(Tasklet):
   def __init__(self, lu, instance_name, cleanup=False,
                failover=False, fallback=False,
                ignore_consistency=False,
-               shutdown_timeout=constants.DEFAULT_SHUTDOWN_TIMEOUT):
+               shutdown_timeout=constants.DEFAULT_SHUTDOWN_TIMEOUT,
+               ignore_ipolicy=False):
     """Initializes this class.
 
     """
@@ -7668,6 +7672,7 @@ class TLMigrateInstance(Tasklet):
     self.fallback = fallback
     self.ignore_consistency = ignore_consistency
     self.shutdown_timeout = shutdown_timeout
+    self.ignore_ipolicy = ignore_ipolicy
 
   def CheckPrereq(self):
     """Check prerequisites.
@@ -7679,6 +7684,7 @@ class TLMigrateInstance(Tasklet):
     instance = self.cfg.GetInstanceInfo(instance_name)
     assert instance is not None
     self.instance = instance
+    cluster = self.cfg.GetClusterInfo()
 
     if (not self.cleanup and
         not instance.admin_state == constants.ADMINST_UP and
@@ -7706,6 +7712,12 @@ class TLMigrateInstance(Tasklet):
         # BuildHooksEnv
         self.target_node = self.lu.op.target_node
 
+      # Check that the target node is correct in terms of instance policy
+      nodeinfo = self.cfg.GetNodeInfo(self.target_node)
+      ipolicy = _CalculateGroupIPolicy(cluster, nodeinfo.group)
+      _CheckTargetNodeIPolicy(self.lu, ipolicy, instance, nodeinfo,
+                              ignore=self.ignore_ipolicy)
+
       # self.target_node is already populated, either directly or by the
       # iallocator run
       target_node = self.target_node
@@ -7739,8 +7751,12 @@ class TLMigrateInstance(Tasklet):
                                    " node can be passed)" %
                                    (instance.disk_template, text),
                                    errors.ECODE_INVAL)
+      nodeinfo = self.cfg.GetNodeInfo(target_node)
+      ipolicy = _CalculateGroupIPolicy(cluster, nodeinfo.group)
+      _CheckTargetNodeIPolicy(self.lu, ipolicy, instance, nodeinfo,
+                              ignore=self.ignore_ipolicy)
 
-    i_be = self.cfg.GetClusterInfo().FillBE(instance)
+    i_be = cluster.FillBE(instance)
 
     # check memory requirements on the secondary node
     if not self.failover or instance.admin_state == constants.ADMINST_UP:
@@ -7796,8 +7812,7 @@ class TLMigrateInstance(Tasklet):
         self.lu.op.live = None
       elif self.lu.op.mode is None:
         # read the default value from the hypervisor
-        i_hv = self.cfg.GetClusterInfo().FillHV(self.instance,
-                                                skip_globals=False)
+        i_hv = cluster.FillHV(self.instance, skip_globals=False)
         self.lu.op.mode = i_hv[constants.HV_MIGRATION_MODE]
 
       self.live = self.lu.op.mode == constants.HT_MIGRATION_LIVE
@@ -7809,6 +7824,7 @@ class TLMigrateInstance(Tasklet):
     """Run the allocator based on input opcode.
 
     """
+    # FIXME: add a self.ignore_ipolicy option
     ial = IAllocator(self.cfg, self.rpc,
                      mode=constants.IALLOCATOR_MODE_RELOC,
                      name=self.instance_name,

diff --git a/lib/opcodes.py b/lib/opcodes.py
index 72ed15d..13074fe 100644 (file)
--- a/lib/opcodes.py
+++ b/lib/opcodes.py
@@ -152,6 +152,10 @@ _PDiskParams = ("diskparams", None,
 _PHvState = ("hv_state", None, ht.TMaybeDict, "Set hypervisor states")
 _PDiskState = ("disk_state", None, ht.TMaybeDict, "Set disk states")
 
+
+_PIgnoreIpolicy = ("ignore_ipolicy", False, ht.TBool,
+                   "Whether to ignore ipolicy violations")
+
 #: OP_ID conversion regular expression
 _OPID_RE = re.compile("([a-z])([A-Z])")
 
@@ -1236,6 +1240,7 @@ class OpInstanceFailover(OpCode):
     _PShutdownTimeout,
     _PIgnoreConsistency,
     _PMigrationTargetNode,
+    _PIgnoreIpolicy,
     ("iallocator", None, ht.TMaybeString,
      "Iallocator for deciding the target node for shared-storage instances"),
     ]

diff --git a/man/gnt-instance.rst b/man/gnt-instance.rst
index 2fab5d4..3363d46 100644 (file)
--- a/man/gnt-instance.rst
+++ b/man/gnt-instance.rst
@@ -1380,7 +1380,7 @@ FAILOVER
 ^^^^^^^^
 
 **failover** [-f] [--ignore-consistency] [--shutdown-timeout=*N*]
-[--submit] {*instance*}
+[--submit] [--ignore-ipolicy] {*instance*}
 
 Failover will stop the instance (if running), change its primary node,
 and if it was originally running it will start it again (on the new
@@ -1406,6 +1406,9 @@ The ``--submit`` option is used to send the job to the master daemon
 but not wait for its completion. The job ID will be shown so that it
 can be examined via **gnt-job info**.
 
+If ``--ignore-ipolicy`` is given any instance policy violations occuring
+during this operation are ignored.
+
 Example::
 
     # gnt-instance failover instance1.example.com