1 gnt-node(8) Ganeti | Version @GANETI_VERSION@
2 =============================================
7 gnt-node - Node administration
12 **gnt-node** {command} [arguments...]
17 The **gnt-node** is used for managing the (physical) nodes in the
26 | **add** [\--readd] [{-s|\--secondary-ip} *secondary\_ip*]
27 | [{-g|\--node-group} *nodegroup*]
28 | [\--master-capable=``yes|no``] [\--vm-capable=``yes|no``]
29 | [\--node-parameters *ndparams*]
30 | [\--disk-state *diskstate*]
31 | [\--hypervisor-state *hvstate*]
34 Adds the given node to the cluster.
36 This command is used to join a new node to the cluster. You will
37 have to provide the password for root of the node to be able to add
38 the node in the cluster. The command needs to be run on the Ganeti
41 Note that the command is potentially destructive, as it will
42 forcibly join the specified host the cluster, not paying attention
43 to its current status (it could be already in a cluster, etc.)
45 The ``-s (--secondary-ip)`` is used in dual-home clusters and
46 specifies the new node's IP in the secondary network. See the
47 discussion in **gnt-cluster**(8) for more information.
49 In case you're readding a node after hardware failure, you can use
50 the ``--readd`` parameter. In this case, you don't need to pass the
51 secondary IP again, it will reused from the cluster. Also, the
52 drained and offline flags of the node will be cleared before
55 The ``-g (--node-group)`` option is used to add the new node into a
56 specific node group, specified by UUID or name. If only one node group
57 exists you can skip this option, otherwise it's mandatory.
59 The ``vm_capable``, ``master_capable``, ``ndparams``, ``diskstate`` and
60 ``hvstate`` options are described in **ganeti**(7), and are used to set
61 the properties of the new node.
63 The command performs some operations that change the state of the master
64 and the new node, like copying certificates and starting the node daemon
65 on the new node, or updating ``/etc/hosts`` on the master node. If the
66 command fails at a later stage, it doesn't undo such changes. This
67 should not be a problem, as a successful run of ``gnt-node add`` will
68 bring everything back in sync.
72 # gnt-node add node5.example.com
73 # gnt-node add -s 192.0.2.5 node5.example.com
74 # gnt-node add -g group2 -s 192.0.2.9 node9.group2.example.com
80 **add-tags** [\--from *file*] {*nodename*} {*tag*...}
82 Add tags to the given node. If any of the tags contains invalid
83 characters, the entire operation will abort.
85 If the ``--from`` option is given, the list of tags will be
86 extended with the contents of that file (each line becomes a tag).
87 In this case, there is not need to pass tags on the command line
88 (if you do, both sources will be used). A file name of - will be
94 | **evacuate** [-f] [\--early-release] [\--submit]
95 | [{-I|\--iallocator} *NAME* \| {-n|\--new-secondary} *destination\_node*]
96 | [{-p|\--primary-only} \| {-s|\--secondary-only} ]
99 This command will move instances away from the given node. If
100 ``--primary-only`` is given, only primary instances are evacuated, with
101 ``--secondary-only`` only secondaries. If neither is given, all
102 instances are evacuated. It works only for instances having a drbd disk
105 The new location for the instances can be specified in two ways:
107 - as a single node for all instances, via the ``-n (--new-secondary)``
110 - or via the ``-I (--iallocator)`` option, giving a script name as
111 parameter (or ``.`` to use the default allocator), so each instance
112 will be in turn placed on the (per the script) optimal node
114 The ``--early-release`` changes the code so that the old storage on
115 node being evacuated is removed early (before the resync is
116 completed) and the internal Ganeti locks are also released for both
117 the current secondary and the new secondary, thus allowing more
118 parallelism in the cluster operation. This should be used only when
119 recovering from a disk failure on the current secondary (thus the
120 old storage is already broken) or when the storage on the primary
121 node is known to be fine (thus we won't need the old storage for
124 Note that this command is equivalent to using per-instance commands for
125 each affected instance individually:
127 - ``--primary-only`` is equivalent to ``gnt-instance
128 failover/migration`` for non-DRBD instances, but for DRBD instances
129 it's different, and usually is a slow process (it will change the
130 primary to another node while keeping the secondary, this requiring
131 data copies, whereas failover/migrate will only toggle the
132 primary/secondary roles, a fast process)
133 - ``--secondary-only`` is equivalent to ``gnt-instance replace-disks``
134 in the secondary node change mode (only valid for DRBD instances)
135 - when neither of the above is done a combination of the two cases is run
137 See **ganeti(7)** for a description of ``--submit`` and other common
142 # gnt-node evacuate -I hail node3.example.com
148 **failover** [-f] [\--ignore-consistency] {*node*}
150 This command will fail over all instances having the given node as
151 primary to their secondary nodes. This works only for instances having
152 a drbd disk template.
154 Normally the failover will check the consistency of the disks before
155 failing over the instance. If you are trying to migrate instances off
156 a dead node, this will fail. Use the ``--ignore-consistency`` option
161 # gnt-node failover node1.example.com
169 Show detailed information about the nodes in the cluster. If you
170 don't give any arguments, all nodes will be shows, otherwise the
171 output will be restricted to the given names.
177 | [\--no-headers] [\--separator=*SEPARATOR*]
178 | [\--units=*UNITS*] [-v] [{-o|\--output} *[+]FIELD,...*]
182 Lists the nodes in the cluster.
184 The ``--no-headers`` option will skip the initial header line. The
185 ``--separator`` option takes an argument which denotes what will be
186 used between the output fields. Both these options are to help
189 The units used to display the numeric values in the output varies,
190 depending on the options given. By default, the values will be
191 formatted in the most appropriate unit. If the ``--separator``
192 option is given, then the values are shown in mebibytes to allow
193 parsing by scripts. In both cases, the ``--units`` option can be
194 used to enforce a given output unit.
196 Queries of nodes will be done in parallel with any running jobs. This might
197 give inconsistent results for the free disk/memory.
199 The ``-v`` option activates verbose mode, which changes the display of
200 special field states (see **ganeti(7)**).
202 The ``-o (--output)`` option takes a comma-separated list of output
203 fields. The available fields and their meaning are:
207 If the value of the option starts with the character ``+``, the new
208 fields will be added to the default list. This allows one to quickly
209 see the default list plus a few other fields, instead of retyping
210 the entire list of fields.
212 Note that some of these fields are known from the configuration of the
213 cluster (e.g. ``name``, ``pinst``, ``sinst``, ``pip``, ``sip``) and thus
214 the master does not need to contact the node for this data (making the
215 listing fast if only fields from this set are selected), whereas the
216 other fields are "live" fields and require a query to the cluster nodes.
218 Depending on the virtualization type and implementation details, the
219 ``mtotal``, ``mnode`` and ``mfree`` fields may have slightly varying
220 meanings. For example, some solutions share the node memory with the
221 pool of memory used for instances (KVM), whereas others have separate
222 memory for the node and for the instances (Xen).
224 If exactly one argument is given and it appears to be a query filter
225 (see **ganeti(7)**), the query result is filtered accordingly. For
226 ambiguous cases (e.g. a single field name as a filter) the ``--filter``
227 (``-F``) option forces the argument to be treated as a filter (e.g.
228 ``gnt-node list -F master_candidate``).
230 If no node names are given, then all nodes are queried. Otherwise,
231 only the given nodes will be listed.
237 **list-drbd** [\--no-headers] [\--separator=*SEPARATOR*] node
239 Lists the mapping of DRBD minors for a given node. This outputs a static
240 list of fields (it doesn't accept the ``--output`` option), as follows:
243 The (full) name of the node we are querying
247 The instance the DRBD minor belongs to
249 The disk index that the DRBD minor belongs to
251 Either ``primary`` or ``secondary``, denoting the role of the node for
252 the instance (note: this is not the live status of the DRBD device,
253 but the configuration value)
255 The node that the minor is connected to on the other end
257 This command can be used as a reverse lookup (from node and minor) to a
258 given instance, which can be useful when debugging DRBD issues.
260 Note that this command queries Ganeti via :manpage:`ganeti-confd(8)`, so
261 it won't be available if support for ``confd`` has not been enabled at
262 build time; furthermore, in Ganeti 2.6 this is only available via the
263 Haskell version of confd (again selected at build time).
268 **list-fields** [field...]
270 Lists available fields for nodes.
276 **list-tags** {*nodename*}
278 List the tags of the given node.
283 | **migrate** [-f] [\--non-live] [\--migration-mode=live\|non-live]
284 | [\--ignore-ipolicy] [\--submit] {*node*}
286 This command will migrate all instances having the given node as
287 primary to their secondary nodes. This works only for instances
288 having a drbd disk template.
290 As for the **gnt-instance migrate** command, the options
291 ``--no-live``, ``--migration-mode`` and ``--no-runtime-changes``
292 can be given to influence the migration type.
294 If ``--ignore-ipolicy`` is given any instance policy violations
295 occurring during this operation are ignored.
297 See **ganeti(7)** for a description of ``--submit`` and other common
302 # gnt-node migrate node1.example.com
308 | **modify** [-f] [\--submit]
309 | [{-C|\--master-candidate} ``yes|no``]
310 | [{-D|\--drained} ``yes|no``] [{-O|\--offline} ``yes|no``]
311 | [\--master-capable=``yes|no``] [\--vm-capable=``yes|no``] [\--auto-promote]
312 | [{-s|\--secondary-ip} *secondary_ip*]
313 | [\--node-parameters *ndparams*]
314 | [\--node-powered=``yes|no``]
315 | [\--hypervisor-state *hvstate*]
316 | [\--disk-state *diskstate*]
319 This command changes the role of the node. Each options takes
320 either a literal yes or no, and only one option should be given as
321 yes. The meaning of the roles and flags are described in the
322 manpage **ganeti(7)**.
324 The option ``--node-powered`` can be used to modify state-of-record if
325 it doesn't reflect the reality anymore.
327 In case a node is demoted from the master candidate role, the
328 operation will be refused unless you pass the ``--auto-promote``
329 option. This option will cause the operation to lock all cluster nodes
330 (thus it will not be able to run in parallel with most other jobs),
331 but it allows automated maintenance of the cluster candidate pool. If
332 locking all cluster node is too expensive, another option is to
333 promote manually another node to master candidate before demoting the
336 Example (setting a node offline, which will demote it from master
337 candidate role if is in that role)::
339 # gnt-node modify --offline=yes node1.example.com
341 The ``-s (--secondary-ip)`` option can be used to change the node's
342 secondary ip. No drbd instances can be running on the node, while this
343 operation is taking place. Remember that the secondary ip must be
344 reachable from the master secondary ip, when being changed, so be sure
345 that the node has the new IP already configured and active. In order to
346 convert a cluster from single homed to multi-homed or vice versa
347 ``--force`` is needed as well, and the target node for the first change
350 See **ganeti(7)** for a description of ``--submit`` and other common
353 Example (setting the node back to online and master candidate)::
355 # gnt-node modify --offline=no --master-candidate=yes node1.example.com
361 **remove** {*nodename*}
363 Removes a node from the cluster. Instances must be removed or
364 migrated to another cluster before.
368 # gnt-node remove node5.example.com
374 **remove-tags** [\--from *file*] {*nodename*} {*tag*...}
376 Remove tags from the given node. If any of the tags are not
377 existing on the node, the entire operation will abort.
379 If the ``--from`` option is given, the list of tags to be removed will
380 be extended with the contents of that file (each line becomes a tag).
381 In this case, there is not need to pass tags on the command line (if
382 you do, tags from both sources will be removed). A file name of - will
383 be interpreted as stdin.
388 | **volumes** [\--no-headers] [\--human-readable]
389 | [\--separator=*SEPARATOR*] [{-o|\--output} *FIELDS*]
392 Lists all logical volumes and their physical disks from the node(s)
395 The ``--no-headers`` option will skip the initial header line. The
396 ``--separator`` option takes an argument which denotes what will be
397 used between the output fields. Both these options are to help
400 The units used to display the numeric values in the output varies,
401 depending on the options given. By default, the values will be
402 formatted in the most appropriate unit. If the ``--separator``
403 option is given, then the values are shown in mebibytes to allow
404 parsing by scripts. In both cases, the ``--units`` option can be
405 used to enforce a given output unit.
407 The ``-o (--output)`` option takes a comma-separated list of output
408 fields. The available fields and their meaning are:
411 the node name on which the volume exists
414 the physical drive (on which the LVM physical volume lives)
417 the volume group name
420 the logical volume name
423 the logical volume size
426 The name of the instance to which this volume belongs, or (in case
427 it's an orphan volume) the character "-"
432 # gnt-node volumes node5.example.com
433 Node PhysDev VG Name Size Instance
434 node1.example.com /dev/hdc1 xenvg instance1.example.com-sda_11000.meta 128 instance1.example.com
435 node1.example.com /dev/hdc1 xenvg instance1.example.com-sda_11001.data 256 instance1.example.com
441 | **list-storage** [\--no-headers] [\--human-readable]
442 | [\--separator=*SEPARATOR*] [\--storage-type=*STORAGE\_TYPE*]
443 | [{-o|\--output} *FIELDS*]
446 Lists the available storage units and their details for the given
449 The ``--no-headers`` option will skip the initial header line. The
450 ``--separator`` option takes an argument which denotes what will be
451 used between the output fields. Both these options are to help
454 The units used to display the numeric values in the output varies,
455 depending on the options given. By default, the values will be
456 formatted in the most appropriate unit. If the ``--separator``
457 option is given, then the values are shown in mebibytes to allow
458 parsing by scripts. In both cases, the ``--units`` option can be
459 used to enforce a given output unit.
461 The ``--storage-type`` option can be used to choose a storage unit
462 type. Possible choices are lvm-pv, lvm-vg or file.
464 The ``-o (--output)`` option takes a comma-separated list of output
465 fields. The available fields and their meaning are:
468 the node name on which the volume exists
471 the type of the storage unit (currently just what is passed in via
475 the path/identifier of the storage unit
478 total size of the unit; for the file type see a note below
481 used space in the unit; for the file type see a note below
487 whether we the unit is available for allocation (only lvm-pv can
488 change this setting, the other types always report true)
491 Note that for the "file" type, the total disk space might not equal
492 to the sum of used and free, due to the method Ganeti uses to
493 compute each of them. The total and free values are computed as the
494 total and free space values for the filesystem to which the
495 directory belongs, but the used space is computed from the used
496 space under that directory *only*, which might not be necessarily
497 the root of the filesystem, and as such there could be files
498 outside the file storage directory using disk space and causing a
499 mismatch in the values.
503 node1# gnt-node list-storage node2
504 Node Type Name Size Used Free Allocatable
505 node2 lvm-pv /dev/sda7 673.8G 1.5G 672.3G Y
506 node2 lvm-pv /dev/sdb1 698.6G 0M 698.6G Y
512 | **modify-storage** [\--allocatable={yes|no}] [\--submit]
513 | {*node*} {*storage-type*} {*volume-name*}
515 Modifies storage volumes on a node. Only LVM physical volumes can
516 be modified at the moment. They have a storage type of "lvm-pv".
520 # gnt-node modify-storage --allocatable no node5.example.com lvm-pv /dev/sdb1
526 | **repair-storage** [\--ignore-consistency] ]\--submit]
527 | {*node*} {*storage-type*} {*volume-name*}
529 Repairs a storage volume on a node. Only LVM volume groups can be
530 repaired at this time. They have the storage type "lvm-vg".
532 On LVM volume groups, **repair-storage** runs ``vgreduce
537 **Caution:** Running this command can lead to data loss. Use it with
540 The ``--ignore-consistency`` option will ignore any inconsistent
541 disks (on the nodes paired with this one). Use of this option is
542 most likely to lead to data-loss.
546 # gnt-node repair-storage node5.example.com lvm-vg xenvg
552 **powercycle** [\--yes] [\--force] [\--submit] {*node*}
554 This command (tries to) forcefully reboot a node. It is a command
555 that can be used if the node environment is broken, such that the
556 admin can no longer login over SSH, but the Ganeti node daemon is
559 Note that this command is not guaranteed to work; it depends on the
560 hypervisor how effective is the reboot attempt. For Linux, this
561 command requires the kernel option ``CONFIG_MAGIC_SYSRQ`` to be
564 The ``--yes`` option can be used to skip confirmation, while the
565 ``--force`` option is needed if the target node is the master
568 See **ganeti(7)** for a description of ``--submit`` and other common
574 **power** [``--force``] [``--ignore-status``] [``--all``]
575 [``--power-delay``] on|off|cycle|status [*nodes*]
577 This command calls out to out-of-band management to change the power
578 state of given node. With ``status`` you get the power status as reported
579 by the out-of-band management script.
581 Note that this command will only work if the out-of-band functionality
582 is configured and enabled on the cluster. If this is not the case,
583 please use the **powercycle** command above.
585 Using ``--force`` you skip the confirmation to do the operation.
586 Currently this only has effect on ``off`` and ``cycle``. On those two
587 you can *not* operate on the master. However, the command will provide
588 you with the command to invoke to operate on the master nerver-mind.
589 This is considered harmful and Ganeti does not support the use of it.
591 Providing ``--ignore-status`` will ignore the offline=N state of a node
592 and continue with power off.
594 ``--power-delay`` specifies the time in seconds (factions allowed)
595 waited between powering on the next node. This is by default 2 seconds
596 but can increased if needed with this option.
598 *nodes* are optional. If not provided it will call out for every node in
599 the cluster. Except for the ``off`` and ``cycle`` command where you've
600 to explicit use ``--all`` to select all.
608 This command calls out to out-of-band management to ask for the health status
609 of all or given nodes. The health contains the node name and then the items
610 element with their status in a ``item=status`` manner. Where ``item`` is script
611 specific and ``status`` can be one of ``OK``, ``WARNING``, ``CRITICAL`` or
612 ``UNKNOWN``. Items with status ``WARNING`` or ``CRITICAL`` are logged and
613 annotated in the command line output.
619 | **restricted-command** [-M] [--sync]
620 | { -g *group* *command* | *command* *nodes*... }
622 Executes a restricted command on the specified nodes. Restricted commands are
623 not arbitrary, but must reside in
624 ``@SYSCONFDIR@/ganeti/restricted-commands`` on a node, either as a regular
625 file or as a symlink. The directory must be owned by root and not be
626 world- or group-writable. If a command fails verification or otherwise
627 fails to start, the node daemon log must be consulted for more detailed
630 Example for running a command on two nodes::
632 # gnt-node restricted-command mycommand \
633 node1.example.com node2.example.com
635 The ``-g`` option can be used to run a command only on a specific node
638 # gnt-node restricted-command -g default mycommand
640 The ``-M`` option can be used to prepend the node name to all command
641 output lines. ``--sync`` forces the opcode to acquire the node lock(s)
644 .. vim: set textwidth=72 :