diff --git a/Makefile.am b/Makefile.am
index b803aaab070db4266984a641d4fca54bd131336b..b90074cac847cc643677f3253e7392af44bf14dc 100644
--- a/Makefile.am
+++ b/Makefile.am
@@ -330,6 +330,7 @@ $(REPLACE_VARS_SED): Makefile stamp-directories
echo 's#@CUSTOM_XEN_KERNEL@#$(XEN_KERNEL)#g'; \
echo 's#@CUSTOM_XEN_INITRD@#$(XEN_INITRD)#g'; \
echo 's#@RPL_FILE_STORAGE_DIR@#$(FILE_STORAGE_DIR)#g'; \
+ echo 's#@PKGLIBDIR@#$(pkglibdir)#g'; \
} > $@
# We need to create symlinks because "make distcheck" will not install Python
diff --git a/NEWS b/NEWS
index cbbf76a78c842add206e4ed73dc8dd07847c047d..f1852058928bc3d3952286c098ae7309067ad30b 100644
--- a/NEWS
+++ b/NEWS
@@ -1,3 +1,15 @@
+Version 2.0.3
+ - Added β--ignore-sizeβ to the βgnt-instance activate-disksβ command
+ to allow using the pre-2.0.2 behaviour in activation, if any
+ existing instances have mismatched disk sizes in the configuration
+ - Added βgnt-cluster repair-disk-sizesβ command to check and update
+ any configuration mismatches for disk sizes
+ - Added βgnt-master cluste-failover --no-votingβ to allow master
+ failover to work on two-node clusters
+ - Fixed the β--netβ option of βgnt-backup importβ, which was unusable
+ - Fixed detection of OS script errors in βgnt-backup exportβ
+ - Fixed exit code of βgnt-backup exportβ
+
Version 2.0.2
- Added experimental support for stripped logical volumes; this should
enhance performance but comes with a higher complexity in the block
diff --git a/configure.ac b/configure.ac
index 12ba06df08c9a98d679eb62c99f6bb6c5bb2166b..6d3b906c1e6da552068b78fe9b05b01cd443ccb9 100644
--- a/configure.ac
+++ b/configure.ac
@@ -1,7 +1,7 @@
# Configure script for Ganeti
m4_define([gnt_version_major], [2])
m4_define([gnt_version_minor], [0])
-m4_define([gnt_version_revision], [2])
+m4_define([gnt_version_revision], [3])
m4_define([gnt_version_suffix], [])
m4_define([gnt_version_full],
m4_format([%d.%d.%d%s],
diff --git a/daemons/ganeti-masterd b/daemons/ganeti-masterd
index d9b8e74021e1afcb6363ac9bd7c49230a9ba891c..51e01ba15b04a60abc8fed3826f3d6472b3a4ca9 100755
--- a/daemons/ganeti-masterd
+++ b/daemons/ganeti-masterd
@@ -524,7 +524,7 @@ def main():
try:
# activate ip
master_node = ssconf.SimpleStore().GetMasterNode()
- if not rpc.RpcRunner.call_node_start_master(master_node, False):
+ if not rpc.RpcRunner.call_node_start_master(master_node, False, False):
logging.error("Can't activate master IP address")
master.setup_queue()
diff --git a/daemons/ganeti-noded b/daemons/ganeti-noded
index fada1fe514cd3ebfd10efa75172816db730d2931..34904724fb9ff9df6144f205a0800dad44d5f2bc 100755
--- a/daemons/ganeti-noded
+++ b/daemons/ganeti-noded
@@ -232,6 +232,14 @@ class NodeHttpServer(http.server.HttpServer):
disks = [objects.Disk.FromDict(cf) for cf in params[1]]
return backend.BlockdevClose(params[0], disks)
+ @staticmethod
+ def perspective_blockdev_getsize(params):
+ """Compute the sizes of the given block devices.
+
+ """
+ disks = [objects.Disk.FromDict(cf) for cf in params[0]]
+ return backend.BlockdevGetsize(disks)
+
# blockdev/drbd specific methods ----------
@staticmethod
@@ -520,7 +528,7 @@ class NodeHttpServer(http.server.HttpServer):
"""Promote this node to master status.
"""
- return backend.StartMaster(params[0])
+ return backend.StartMaster(params[0], params[1])
@staticmethod
def perspective_node_stop_master(params):
diff --git a/devel/upload.in b/devel/upload.in
index 94759e60f9c7b8276c4ccb583aae446b7cf334c4..733befa338d9e9d7c50a9340297fbeab602cb942 100644
--- a/devel/upload.in
+++ b/devel/upload.in
@@ -27,14 +27,22 @@
set -e
+PREFIX='@PREFIX@'
+SYSCONFDIR='@SYSCONFDIR@'
+PKGLIBDIR='@PKGLIBDIR@'
+
NO_RESTART=
+NO_CRON=
hosts=
while [ "$#" -gt 0 ]; do
opt="$1"
case "$opt" in
--no-restart)
NO_RESTART=1
- ;;
+ ;;
+ --no-cron)
+ NO_CRON=1
+ ;;
-h|--help)
echo "Usage: $0 [--no-restart] hosts..."
exit 0
@@ -42,10 +50,10 @@ while [ "$#" -gt 0 ]; do
-*)
echo "Unknown option: $opt" >&2
exit 1
- ;;
+ ;;
*)
hosts="$hosts $opt"
- ;;
+ ;;
esac
shift
done
@@ -58,39 +66,36 @@ trap 'rm -rf $TXD' EXIT
# install ganeti as a real tree
make install DESTDIR="$TXD"
+# copy additional needed files
+install -D --mode=0755 doc/examples/ganeti.initd \
+ "$TXD/$SYSCONFDIR/init.d/ganeti"
+
+install -D --mode=0644 doc/examples/bash_completion \
+ "$TXD/$SYSCONFDIR/bash_completion.d/ganeti"
+
+if [ -z "$NO_CRON" ]; then
+ install -D --mode=0644 doc/examples/ganeti.cron \
+ "$TXD/$SYSCONFDIR/cron.d/ganeti"
+fi
+
+install -D --mode=0755 doc/examples/dumb-allocator \
+ "$TXD/$PKGLIBDIR/iallocators/dumb"
+
echo ---
( cd "$TXD" && find; )
echo ---
-PREFIX='@PREFIX@'
-
# and now put it under $prefix on the target node(s)
for host; do
echo Uploading code to ${host}...
rsync -v -rlDc --exclude="*.py[oc]" --exclude="*.pdf" --exclude="*.html" \
- "$TXD/$PREFIX/" \
- root@${host}:$PREFIX/ &
-done
-wait
-
-INIT_SCRIPT="$TXD/ganeti.initd"
-install --mode=0755 doc/examples/ganeti.initd $INIT_SCRIPT
-for host; do
- echo Uploading init script to ${host}...
- scp $INIT_SCRIPT root@${host}:/etc/init.d/ganeti &
+ "$TXD/" \
+ root@${host}:/ &
done
wait
-if [ -f ganeti-master-cron ]; then
- for host; do
- echo Uploading cron files to ${host}...
- scp ganeti-master-cron root@${host}:/etc/ganeti/master-cron &
- done
-fi
-wait
-
if test -z "${NO_RESTART}"; then
for host; do
echo Restarting ganeti-noded on ${host}...
diff --git a/doc/examples/bash_completion.in b/doc/examples/bash_completion.in
index 09fe0eafb0ee9989854b2ddbdd562dabccf2dd5f..b52c10d6c280d7f07cefbedc462c5848add2450b 100644
--- a/doc/examples/bash_completion.in
+++ b/doc/examples/bash_completion.in
@@ -90,7 +90,7 @@ _gnt_cluster()
if [[ -e "@LOCALSTATEDIR@/lib/ganeti/ssconf_cluster_name" ]]; then
cmds="add-tags command copyfile destroy getmaster info list-tags \
masterfailover modify queue redist-conf remove-tags rename \
- search-tags verify verify-disks version"
+ repair-disk-sizes search-tags verify verify-disks version"
else
cmds="init"
fi
diff --git a/lib/backend.py b/lib/backend.py
index dbc32f6f7b67cc38417f96fb5ac00161fd725db3..f4e308e651ad505818c89543ee98c202863ba5e1 100644
--- a/lib/backend.py
+++ b/lib/backend.py
@@ -169,7 +169,7 @@ def GetMasterInfo():
return (master_netdev, master_ip, master_node)
-def StartMaster(start_daemons):
+def StartMaster(start_daemons, no_voting):
"""Activate local node as master node.
The function will always try activate the IP address of the master
@@ -179,6 +179,9 @@ def StartMaster(start_daemons):
@type start_daemons: boolean
@param start_daemons: whther to also start the master
daemons (ganeti-masterd and ganeti-rapi)
+ @type no_voting: boolean
+ @param no_voting: whether to start ganeti-masterd without a node vote
+ (if start_daemons is True), but still non-interactively
@rtype: None
"""
@@ -208,8 +211,17 @@ def StartMaster(start_daemons):
# and now start the master and rapi daemons
if start_daemons:
- for daemon in 'ganeti-masterd', 'ganeti-rapi':
- result = utils.RunCmd([daemon])
+ daemons_params = {
+ 'ganeti-masterd': [],
+ 'ganeti-rapi': [],
+ }
+ if no_voting:
+ daemons_params['ganeti-masterd'].append('--no-voting')
+ daemons_params['ganeti-masterd'].append('--yes-do-it')
+ for daemon in daemons_params:
+ cmd = [daemon]
+ cmd.extend(daemons_params[daemon])
+ result = utils.RunCmd(cmd)
if result.failed:
logging.error("Can't start daemon %s: %s", daemon, result.output)
ok = False
@@ -1452,6 +1464,32 @@ def BlockdevFind(disk):
return (True, (rbd.dev_path, rbd.major, rbd.minor) + rbd.GetSyncStatus())
+def BlockdevGetsize(disks):
+ """Computes the size of the given disks.
+
+ If a disk is not found, returns None instead.
+
+ @type disks: list of L{objects.Disk}
+ @param disks: the list of disk to compute the size for
+ @rtype: list
+ @return: list with elements None if the disk cannot be found,
+ otherwise the size
+
+ """
+ result = []
+ for cf in disks:
+ try:
+ rbd = _RecursiveFindBD(cf)
+ except errors.BlockDeviceError, err:
+ result.append(None)
+ continue
+ if rbd is None:
+ result.append(None)
+ else:
+ result.append(rbd.GetActualSize())
+ return result
+
+
def UploadFile(file_name, data, mode, uid, gid, atime, mtime):
"""Write a file to the filesystem.
@@ -1815,8 +1853,8 @@ def ExportSnapshot(disk, dest_node, instance, cluster_name, idx):
# the target command is built out of three individual commands,
# which are joined by pipes; we check each individual command for
# valid parameters
- expcmd = utils.BuildShellCmd("cd %s; %s 2>%s", inst_os.path,
- export_script, logfile)
+ expcmd = utils.BuildShellCmd("set -e; set -o pipefail; cd %s; %s 2>%s",
+ inst_os.path, export_script, logfile)
comprcmd = "gzip"
@@ -1829,7 +1867,7 @@ def ExportSnapshot(disk, dest_node, instance, cluster_name, idx):
# all commands have been checked, so we're safe to combine them
command = '|'.join([expcmd, comprcmd, utils.ShellQuoteArgs(remotecmd)])
- result = utils.RunCmd(command, env=export_env)
+ result = utils.RunCmd(["bash", "-c", command], env=export_env)
if result.failed:
logging.error("os snapshot export command '%s' returned error: %s"
diff --git a/lib/bdev.py b/lib/bdev.py
index 9ac41e50e6db08d3b7d99c5bf0d2ed185f5825cc..7055ccb01262f76afa34bad6f50f65d3d5527863 100644
--- a/lib/bdev.py
+++ b/lib/bdev.py
@@ -277,6 +277,23 @@ class BlockDev(object):
"""
raise NotImplementedError
+ def GetActualSize(self):
+ """Return the actual disk size.
+
+ @note: the device needs to be active when this is called
+
+ """
+ assert self.attached, "BlockDevice not attached in GetActualSize()"
+ result = utils.RunCmd(["blockdev", "--getsize64", self.dev_path])
+ if result.failed:
+ _ThrowError("blockdev failed (%s): %s",
+ result.fail_reason, result.output)
+ try:
+ sz = int(result.output.strip())
+ except (ValueError, TypeError), err:
+ _ThrowError("Failed to parse blockdev output: %s", str(err))
+ return sz
+
def __repr__(self):
return ("<%s: unique_id: %s, children: %s, %s:%s, %s>" %
(self.__class__, self.unique_id, self._children,
@@ -1129,9 +1146,10 @@ class DRBD8(BaseDRBD):
"""
args = ["drbdsetup", cls._DevPath(minor), "disk",
backend, meta, "0",
- "-d", "%sm" % size,
"-e", "detach",
"--create-device"]
+ if size:
+ args.extend(["-d", "%sm" % size])
result = utils.RunCmd(args)
if result.failed:
_ThrowError("drbd%d: can't attach local disk: %s", minor, result.output)
@@ -1727,6 +1745,19 @@ class FileStorage(BlockDev):
self.attached = os.path.exists(self.dev_path)
return self.attached
+ def GetActualSize(self):
+ """Return the actual disk size.
+
+ @note: the device needs to be active when this is called
+
+ """
+ assert self.attached, "BlockDevice not attached in GetActualSize()"
+ try:
+ st = os.stat(self.dev_path)
+ return st.st_size
+ except OSError, err:
+ _ThrowError("Can't stat %s: %s", self.dev_path, err)
+
@classmethod
def Create(cls, unique_id, children, size):
"""Create a new file.
diff --git a/lib/bootstrap.py b/lib/bootstrap.py
index 7cd0d92ef9ddb97a37c774f47f56883fa7dee8d4..fda1c2ee64dc0f7b711c2c2c6668fcee637c5762 100644
--- a/lib/bootstrap.py
+++ b/lib/bootstrap.py
@@ -275,7 +275,7 @@ def InitCluster(cluster_name, mac_prefix, def_bridge,
# start the master ip
# TODO: Review rpc call from bootstrap
- rpc.RpcRunner.call_node_start_master(hostname.name, True)
+ rpc.RpcRunner.call_node_start_master(hostname.name, True, False)
def InitConfig(version, cluster_config, master_node_config,
@@ -453,8 +453,7 @@ def MasterFailover(no_voting=False):
# cluster info
cfg.Update(cluster_info)
- # 2.0.X: Don't start the master if no_voting is true
- result = rpc.RpcRunner.call_node_start_master(new_master, not no_voting)
+ result = rpc.RpcRunner.call_node_start_master(new_master, True, no_voting)
if result.failed or not result.data:
logging.error("Could not start the master role on the new master"
" %s, please check", new_master)
diff --git a/lib/cmdlib.py b/lib/cmdlib.py
index 67535363aad147bee70009965d4a9456559f8145..ae2267ad560c75fb0cd57b47bf1ea02eddbbdf32 100644
--- a/lib/cmdlib.py
+++ b/lib/cmdlib.py
@@ -1325,6 +1325,100 @@ class LUVerifyDisks(NoHooksLU):
return result
+class LURepairDiskSizes(NoHooksLU):
+ """Verifies the cluster disks sizes.
+
+ """
+ _OP_REQP = ["instances"]
+ REQ_BGL = False
+
+ def ExpandNames(self):
+
+ if not isinstance(self.op.instances, list):
+ raise errors.OpPrereqError("Invalid argument type 'instances'")
+
+ if self.op.instances:
+ self.wanted_names = []
+ for name in self.op.instances:
+ full_name = self.cfg.ExpandInstanceName(name)
+ if full_name is None:
+ raise errors.OpPrereqError("Instance '%s' not known" % name)
+ self.wanted_names.append(full_name)
+ self.needed_locks[locking.LEVEL_INSTANCE] = self.wanted_names
+ self.needed_locks = {
+ locking.LEVEL_NODE: [],
+ locking.LEVEL_INSTANCE: self.wanted_names,
+ }
+ self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
+ else:
+ self.wanted_names = None
+ self.needed_locks = {
+ locking.LEVEL_NODE: locking.ALL_SET,
+ locking.LEVEL_INSTANCE: locking.ALL_SET,
+ }
+ self.share_locks = dict(((i, 1) for i in locking.LEVELS))
+
+ def DeclareLocks(self, level):
+ if level == locking.LEVEL_NODE and self.wanted_names is not None:
+ self._LockInstancesNodes(primary_only=True)
+
+ def CheckPrereq(self):
+ """Check prerequisites.
+
+ This only checks the optional instance list against the existing names.
+
+ """
+ if self.wanted_names is None:
+ self.wanted_names = self.acquired_locks[locking.LEVEL_INSTANCE]
+
+ self.wanted_instances = [self.cfg.GetInstanceInfo(name) for name
+ in self.wanted_names]
+
+ def Exec(self, feedback_fn):
+ """Verify the size of cluster disks.
+
+ """
+ # TODO: check child disks too
+ # TODO: check differences in size between primary/secondary nodes
+ per_node_disks = {}
+ for instance in self.wanted_instances:
+ pnode = instance.primary_node
+ if pnode not in per_node_disks:
+ per_node_disks[pnode] = []
+ for idx, disk in enumerate(instance.disks):
+ per_node_disks[pnode].append((instance, idx, disk))
+
+ changed = []
+ for node, dskl in per_node_disks.items():
+ result = self.rpc.call_blockdev_getsizes(node, [v[2] for v in dskl])
+ if result.failed:
+ self.LogWarning("Failure in blockdev_getsizes call to node"
+ " %s, ignoring", node)
+ continue
+ if len(result.data) != len(dskl):
+ self.LogWarning("Invalid result from node %s, ignoring node results",
+ node)
+ continue
+ for ((instance, idx, disk), size) in zip(dskl, result.data):
+ if size is None:
+ self.LogWarning("Disk %d of instance %s did not return size"
+ " information, ignoring", idx, instance.name)
+ continue
+ if not isinstance(size, (int, long)):
+ self.LogWarning("Disk %d of instance %s did not return valid"
+ " size information, ignoring", idx, instance.name)
+ continue
+ size = size >> 20
+ if size != disk.size:
+ self.LogInfo("Disk %d of instance %s has mismatched size,"
+ " correcting: recorded %d, actual %d", idx,
+ instance.name, disk.size, size)
+ disk.size = size
+ self.cfg.Update(instance)
+ changed.append((instance.name, idx, size))
+ return changed
+
+
class LURenameCluster(LogicalUnit):
"""Rename the cluster.
@@ -1399,7 +1493,7 @@ class LURenameCluster(LogicalUnit):
constants.SSH_KNOWN_HOSTS_FILE, to_node)
finally:
- result = self.rpc.call_node_start_master(master, False)
+ result = self.rpc.call_node_start_master(master, False, False)
if result.failed or not result.data:
self.LogWarning("Could not re-enable the master role on"
" the master, please restart manually.")
@@ -2595,19 +2689,24 @@ class LUActivateInstanceDisks(NoHooksLU):
assert self.instance is not None, \
"Cannot retrieve locked instance %s" % self.op.instance_name
_CheckNodeOnline(self, self.instance.primary_node)
+ if not hasattr(self.op, "ignore_size"):
+ self.op.ignore_size = False
def Exec(self, feedback_fn):
"""Activate the disks.
"""
- disks_ok, disks_info = _AssembleInstanceDisks(self, self.instance)
+ disks_ok, disks_info = \
+ _AssembleInstanceDisks(self, self.instance,
+ ignore_size=self.op.ignore_size)
if not disks_ok:
raise errors.OpExecError("Cannot activate block devices")
return disks_info
-def _AssembleInstanceDisks(lu, instance, ignore_secondaries=False):
+def _AssembleInstanceDisks(lu, instance, ignore_secondaries=False,
+ ignore_size=False):
"""Prepare the block devices for an instance.
This sets up the block devices on all nodes.
@@ -2619,6 +2718,10 @@ def _AssembleInstanceDisks(lu, instance, ignore_secondaries=False):
@type ignore_secondaries: boolean
@param ignore_secondaries: if true, errors on secondary nodes
won't result in an error return from the function
+ @type ignore_size: boolean
+ @param ignore_size: if true, the current known size of the disk
+ will not be used during the disk activation, useful for cases
+ when the size is wrong
@return: False if the operation failed, otherwise a list of
(host, instance_visible_name, node_visible_name)
with the mapping from node devices to instance devices
@@ -2639,6 +2742,9 @@ def _AssembleInstanceDisks(lu, instance, ignore_secondaries=False):
# 1st pass, assemble on all nodes in secondary mode
for inst_disk in instance.disks:
for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node):
+ if ignore_size:
+ node_disk = node_disk.Copy()
+ node_disk.UnsetSize()
lu.cfg.SetDiskID(node_disk, node)
result = lu.rpc.call_blockdev_assemble(node, node_disk, iname, False)
msg = result.RemoteFailMsg()
@@ -2656,6 +2762,9 @@ def _AssembleInstanceDisks(lu, instance, ignore_secondaries=False):
for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node):
if node != instance.primary_node:
continue
+ if ignore_size:
+ node_disk = node_disk.Copy()
+ node_disk.UnsetSize()
lu.cfg.SetDiskID(node_disk, node)
result = lu.rpc.call_blockdev_assemble(node, node_disk, iname, True)
msg = result.RemoteFailMsg()
@@ -6365,6 +6474,8 @@ class LUExportInstance(LogicalUnit):
for disk in instance.disks:
self.cfg.SetDiskID(disk, src_node)
+ # per-disk results
+ dresults = []
try:
for idx, disk in enumerate(instance.disks):
# new_dev_name will be a snapshot of an lvm leaf of the one we passed
@@ -6398,15 +6509,22 @@ class LUExportInstance(LogicalUnit):
if result.failed or not result.data:
self.LogWarning("Could not export disk/%d from node %s to"
" node %s", idx, src_node, dst_node.name)
+ dresults.append(False)
+ else:
+ dresults.append(True)
msg = self.rpc.call_blockdev_remove(src_node, dev).RemoteFailMsg()
if msg:
self.LogWarning("Could not remove snapshot for disk/%d from node"
" %s: %s", idx, src_node, msg)
+ else:
+ dresults.append(False)
result = self.rpc.call_finalize_export(dst_node.name, instance, snap_disks)
+ fin_resu = True
if result.failed or not result.data:
self.LogWarning("Could not finalize export for instance %s on node %s",
instance.name, dst_node.name)
+ fin_resu = False
nodelist = self.cfg.GetNodeList()
nodelist.remove(dst_node.name)
@@ -6423,6 +6541,7 @@ class LUExportInstance(LogicalUnit):
if not self.rpc.call_export_remove(node, instance.name):
self.LogWarning("Could not remove older export for instance %s"
" on node %s", instance.name, node)
+ return fin_resu, dresults
class LURemoveExport(NoHooksLU):
diff --git a/lib/mcpu.py b/lib/mcpu.py
index 67c5e95770bd0f1741586fcac85fe869484c09d4..959a83785331c847569a3f2f0c083f838a76cdf9 100644
--- a/lib/mcpu.py
+++ b/lib/mcpu.py
@@ -50,6 +50,7 @@ class Processor(object):
opcodes.OpVerifyDisks: cmdlib.LUVerifyDisks,
opcodes.OpSetClusterParams: cmdlib.LUSetClusterParams,
opcodes.OpRedistributeConfig: cmdlib.LURedistributeConfig,
+ opcodes.OpRepairDiskSizes: cmdlib.LURepairDiskSizes,
# node lu
opcodes.OpAddNode: cmdlib.LUAddNode,
opcodes.OpQueryNodes: cmdlib.LUQueryNodes,
diff --git a/lib/objects.py b/lib/objects.py
index 5c2ff3d1aeefa64c7c9506da6e0ba524535e635a..acb24d84b6c6f0654dfc045668b35af5df012052 100644
--- a/lib/objects.py
+++ b/lib/objects.py
@@ -153,6 +153,14 @@ class ConfigObject(object):
" _ContainerFromDicts" % c_type)
return ret
+ def Copy(self):
+ """Makes a deep copy of the current object and its children.
+
+ """
+ dict_form = self.ToDict()
+ clone_obj = self.__class__.FromDict(dict_form)
+ return clone_obj
+
def __repr__(self):
"""Implement __repr__ for ConfigObjects."""
return repr(self.ToDict())
@@ -388,6 +396,15 @@ class Disk(ConfigObject):
raise errors.ProgrammerError("Disk.RecordGrow called for unsupported"
" disk type %s" % self.dev_type)
+ def UnsetSize(self):
+ """Sets recursively the size to zero for the disk and its children.
+
+ """
+ if self.children:
+ for child in self.children:
+ child.UnsetSize()
+ self.size = 0
+
def SetPhysicalID(self, target_node, nodes_ip):
"""Convert the logical ID to the physical ID.
diff --git a/lib/opcodes.py b/lib/opcodes.py
index 6ba30f614e2adde716d4f0642e869c05c28f4f2e..6a86477adfdc7163672e6952c47bc3433be5102a 100644
--- a/lib/opcodes.py
+++ b/lib/opcodes.py
@@ -224,6 +224,26 @@ class OpVerifyDisks(OpCode):
__slots__ = []
+class OpRepairDiskSizes(OpCode):
+ """Verify the disk sizes of the instances and fixes configuration
+ mimatches.
+
+ Parameters: optional instances list, in case we want to restrict the
+ checks to only a subset of the instances.
+
+ Result: a list of tuples, (instance, disk, new-size) for changed
+ configurations.
+
+ In normal operation, the list should be empty.
+
+ @type instances: list
+ @ivar instances: the list of instances to check, or empty for all instances
+
+ """
+ OP_ID = "OP_CLUSTER_REPAIR_DISK_SIZES"
+ __slots__ = ["instances"]
+
+
class OpQueryConfigValues(OpCode):
"""Query cluster configuration values."""
OP_ID = "OP_CLUSTER_CONFIG_QUERY"
@@ -433,7 +453,7 @@ class OpActivateInstanceDisks(OpCode):
"""Activate an instance's disks."""
OP_ID = "OP_INSTANCE_ACTIVATE_DISKS"
OP_DSC_FIELD = "instance_name"
- __slots__ = ["instance_name"]
+ __slots__ = ["instance_name", "ignore_size"]
class OpDeactivateInstanceDisks(OpCode):
diff --git a/lib/rpc.py b/lib/rpc.py
index b654f6f1d1c6c63e96be88b07affecf000d8a818..e13885903c597c7ffdab6b252ad8efdec48a6750 100644
--- a/lib/rpc.py
+++ b/lib/rpc.py
@@ -681,14 +681,14 @@ class RpcRunner(object):
[checkdict, cluster_name])
@classmethod
- def call_node_start_master(cls, node, start_daemons):
+ def call_node_start_master(cls, node, start_daemons, no_voting):
"""Tells a node to activate itself as a master.
This is a single-node call.
"""
return cls._StaticSingleNodeCall(node, "node_start_master",
- [start_daemons])
+ [start_daemons, no_voting])
@classmethod
def call_node_stop_master(cls, node, stop_daemons):
@@ -806,6 +806,15 @@ class RpcRunner(object):
params = [instance_name, [cf.ToDict() for cf in disks]]
return self._SingleNodeCall(node, "blockdev_close", params)
+ def call_blockdev_getsizes(self, node, disks):
+ """Returns the size of the given disks.
+
+ This is a single-node call.
+
+ """
+ params = [[cf.ToDict() for cf in disks]]
+ return self._SingleNodeCall(node, "blockdev_getsize", params)
+
def call_drbd_disconnect_net(self, node_list, nodes_ip, disks):
"""Disconnects the network of the given drbd devices.
diff --git a/man/gnt-backup.sgml b/man/gnt-backup.sgml
index d432b6ef6eda9190005f9752ce3df65579ea7c5c..ae3122ac661d4892b0c2064e38177e55b94c06e2 100644
--- a/man/gnt-backup.sgml
+++ b/man/gnt-backup.sgml
@@ -82,6 +82,16 @@
in the exported dump.
</para>
+ <para>
+ The exit code of the command is 0 if all disks were backed up
+ successfully, 1 if no data was backed up or if the
+ configuration export failed, and 2 if just some of the disks
+ failed to backup. The exact details of the failures will be
+ shown during the command execution (and will be stored in the
+ job log). It is recommended that for any non-zero exit code,
+ the backup is considered invalid, and retried.
+ </para>
+
<para>
Example:
<screen>
diff --git a/man/gnt-cluster.sgml b/man/gnt-cluster.sgml
index 98ec7db76eb603337294d680669b7b7991f990f4..b7fecf0bef6ff76fb133e98bdbf8d3a55e5b2945 100644
--- a/man/gnt-cluster.sgml
+++ b/man/gnt-cluster.sgml
@@ -460,14 +460,6 @@
and gnt-cluster redist-conf to make sure the cluster is consistent again.
</para>
- <para>
- In version 2.0.X ganeti-masterd will not be able to start if
- masterfailover is called with the --no-voting option (which, again,
- should only be used on 2 nodes clusters with the former master being
- down). In that case just start it manually passing --no-voting to it
- as well, until you have restored cluster redundancy.
- </para>
-
</refsect2>
<refsect2>
@@ -618,6 +610,38 @@
</para>
</refsect2>
+ <refsect2>
+ <title>REPAIR-DISK-SIZES</title>
+
+ <cmdsynopsis>
+ <command>repair-disk-sizes</command>
+ <arg rep="repeat">instance</arg>
+ </cmdsynopsis>
+
+ <para>
+ This command checks that the recorded size of the given
+ instance's disks matches the actual size and updates any
+ mismatches found. This is needed if the Ganeti configuration
+ is no longer consistent with reality, as it will impact some
+ disk operations. If no arguments are given, all instances will
+ be checked.
+ </para>
+
+ <para>
+ Note that only active disks can be checked by this command; in
+ case a disk cannot be activated it's advised to use
+ <command>gnt-instance activate-disks --ignore-size
+ ...</command> to force activation without regard to the
+ current size.
+ </para>
+
+ <para>
+ When the all disk sizes are consistent, the command will
+ return no output. Otherwise it will log details about the
+ inconsistencies in the configuration.
+ </para>
+ </refsect2>
+
<refsect2>
<title>SEARCH-TAGS</title>
diff --git a/man/gnt-instance.sgml b/man/gnt-instance.sgml
index c4dd54df32d8ae4ff199bdfbbb9f7195fb858ebf..a703d68b6a21136284a21b52f7043328a7ca3374 100644
--- a/man/gnt-instance.sgml
+++ b/man/gnt-instance.sgml
@@ -1673,6 +1673,7 @@ instance5: 11225
<cmdsynopsis>
<command>activate-disks</command>
<arg>--submit</arg>
+ <arg>--ignore-size</arg>
<arg choice="req"><replaceable>instance</replaceable></arg>
</cmdsynopsis>
<para>
@@ -1701,6 +1702,16 @@ node1.example.com:disk/1:/dev/drbd1
<command>gnt-job info</command>.
</para>
+ <para>
+ The <option>--ignore-size</option> option can be used to
+ activate disks ignoring the currently configured size in
+ Ganeti. This can be used in cases where the configuration
+ has gotten out of sync with the real-world (e.g. after a
+ partially-failed grow-disk operation or due to rounding in
+ LVM devices). This should not be used in normal cases, but
+ only when activate-disks fails without it.
+ </para>
+
<para>
Note that it is safe to run this command while the instance
is already running.
diff --git a/scripts/gnt-backup b/scripts/gnt-backup
index e78abbaee05178a57808167597523e47a324d0c0..cbbde51f4bfff6fb58b031f05ac40c6830444327 100755
--- a/scripts/gnt-backup
+++ b/scripts/gnt-backup
@@ -74,8 +74,27 @@ def ExportInstance(opts, args):
target_node=opts.node,
shutdown=opts.shutdown)
- SubmitOpCode(op)
-
+ fin_resu, dlist = SubmitOpCode(op)
+ if not isinstance(dlist, list):
+ ToStderr("Cannot parse execution results")
+ return 1
+ tot_dsk = len(dlist)
+ # TODO: handle diskless instances
+ if dlist.count(False) == 0:
+ # all OK
+ rcode = 0
+ elif dlist.count(True) == 0:
+ ToStderr("Error: No disks were backed up successfully."
+ " The export doesn't have any valid data,"
+ " it is recommended to retry the operation.")
+ rcode = 1
+ else:
+ ToStderr("Partial export failure: %d disks backed up, %d disks failed.",
+ dlist.count(True), dlist.count(False))
+ rcode = 2
+ if not fin_resu:
+ rcode = 1
+ return rcode
def ImportInstance(opts, args):
"""Add an instance to the cluster.
@@ -102,7 +121,7 @@ def ImportInstance(opts, args):
except ValueError, err:
raise errors.OpPrereqError("Invalid NIC index passed: %s" % str(err))
nics = [{}] * nic_max
- for nidx, ndict in opts.nics.items():
+ for nidx, ndict in opts.nics:
nidx = int(nidx)
nics[nidx] = ndict
elif opts.no_nics:
diff --git a/scripts/gnt-cluster b/scripts/gnt-cluster
index f1541670509a24afaf6f285d696440dff9161c56..e6cb4f70cbf4cfb0745dd928f504d76cde04614e 100755
--- a/scripts/gnt-cluster
+++ b/scripts/gnt-cluster
@@ -404,6 +404,20 @@ def VerifyDisks(opts, args):
return retcode
+def RepairDiskSizes(opts, args):
+ """Verify sizes of cluster disks.
+
+ @param opts: the command line options selected by the user
+ @type args: list
+ @param args: optional list of instances to restrict check to
+ @rtype: int
+ @return: the desired exit code
+
+ """
+ op = opcodes.OpRepairDiskSizes(instances=args)
+ SubmitOpCode(op)
+
+
@UsesRPC
def MasterFailover(opts, args):
"""Failover the master node.
@@ -616,6 +630,8 @@ commands = {
"", "Does a check on the cluster configuration"),
'verify-disks': (VerifyDisks, ARGS_NONE, [DEBUG_OPT],
"", "Does a check on the cluster disk status"),
+ 'repair-disk-sizes': (RepairDiskSizes, ARGS_ANY, [DEBUG_OPT],
+ "", "Updates mismatches in recorded disk sizes"),
'masterfailover': (MasterFailover, ARGS_NONE, [DEBUG_OPT,
make_option("--no-voting", dest="no_voting",
help="Skip node agreement check (dangerous)",
diff --git a/scripts/gnt-instance b/scripts/gnt-instance
index 3e8ce3f358ef52d88ca523959effa79eb2308cab..056337700f6d61bc503d623bfca802e2baf98373 100755
--- a/scripts/gnt-instance
+++ b/scripts/gnt-instance
@@ -632,7 +632,8 @@ def ActivateDisks(opts, args):
"""
instance_name = args[0]
- op = opcodes.OpActivateInstanceDisks(instance_name=instance_name)
+ op = opcodes.OpActivateInstanceDisks(instance_name=instance_name,
+ ignore_size=opts.ignore_size)
disks_info = SubmitOrSend(op, opts)
for host, iname, nname in disks_info:
ToStdout("%s:%s:%s", host, iname, nname)
@@ -1508,7 +1509,14 @@ commands = {
SUBMIT_OPT,
],
"<instance>", "Reboots an instance"),
- 'activate-disks': (ActivateDisks, ARGS_ONE, [DEBUG_OPT, SUBMIT_OPT],
+ 'activate-disks': (ActivateDisks, ARGS_ONE,
+ [DEBUG_OPT, SUBMIT_OPT,
+ make_option("--ignore-size", dest="ignore_size",
+ default=False, action="store_true",
+ help="Ignore current recorded size"
+ " (useful for forcing activation when"
+ " the recorded size is wrong)"),
+ ],
"<instance>",
"Activate an instance's disks"),
'deactivate-disks': (DeactivateDisks, ARGS_ONE, [DEBUG_OPT, SUBMIT_OPT],