Synnefo » snf-ganeti » ganeti-local

Ganeti administrator's guide

============================

Documents Ganeti version |version|

.. contents::

.. highlight:: shell-example

Introduction

------------

Ganeti is a virtualization cluster management software. You are expected

to be a system administrator familiar with your Linux distribution and

the Xen or KVM virtualization environments before using it.

The various components of Ganeti all have man pages and interactive

help. This manual though will help you getting familiar with the system

by explaining the most common operations, grouped by related use.

After a terminology glossary and a section on the prerequisites needed

to use this manual, the rest of this document is divided in sections

for the different targets that a command affects: instance, nodes, etc.

.. _terminology-label:

Ganeti terminology

++++++++++++++++++

This section provides a small introduction to Ganeti terminology, which

might be useful when reading the rest of the document.

Cluster

~~~~~~~

A set of machines (nodes) that cooperate to offer a coherent, highly

available virtualization service under a single administration domain.

Node

~~~~

A physical machine which is member of a cluster.  Nodes are the basic

cluster infrastructure, and they don't need to be fault tolerant in

order to achieve high availability for instances.

Node can be added and removed (if they host no instances) at will from

the cluster. In a HA cluster and only with HA instances, the loss of any

single node will not cause disk data loss for any instance; of course,

a node crash will cause the crash of its primary instances.

A node belonging to a cluster can be in one of the following roles at a

given time:

- *master* node, which is the node from which the cluster is controlled

- *master candidate* node, only nodes in this role have the full cluster

  configuration and knowledge, and only master candidates can become the

  master node

- *regular* node, which is the state in which most nodes will be on

  bigger clusters (>20 nodes)

- *drained* node, nodes in this state are functioning normally but the

  cannot receive new instances; the intention is that nodes in this role

  have some issue and they are being evacuated for hardware repairs

- *offline* node, in which there is a record in the cluster

  configuration about the node, but the daemons on the master node will

  not talk to this node; any instances declared as having an offline

  node as either primary or secondary will be flagged as an error in the

  cluster verify operation

Depending on the role, each node will run a set of daemons:

- the :command:`ganeti-noded` daemon, which controls the manipulation of

  this node's hardware resources; it runs on all nodes which are in a

  cluster

- the :command:`ganeti-confd` daemon (Ganeti 2.1+) which runs on all

  nodes, but is only functional on master candidate nodes; this daemon

  can be disabled at configuration time if you don't need its

  functionality

- the :command:`ganeti-rapi` daemon which runs on the master node and

  offers an HTTP-based API for the cluster

- the :command:`ganeti-masterd` daemon which runs on the master node and

  allows control of the cluster

Beside the node role, there are other node flags that influence its

behaviour:

- the *master_capable* flag denotes whether the node can ever become a

  master candidate; setting this to 'no' means that auto-promotion will

  never make this node a master candidate; this flag can be useful for a

  remote node that only runs local instances, and having it become a

  master is impractical due to networking or other constraints

- the *vm_capable* flag denotes whether the node can host instances or

  not; for example, one might use a non-vm_capable node just as a master

  candidate, for configuration backups; setting this flag to no

  disallows placement of instances of this node, deactivates hypervisor

  and related checks on it (e.g. bridge checks, LVM check, etc.), and

  removes it from cluster capacity computations

Instance

~~~~~~~~

A virtual machine which runs on a cluster. It can be a fault tolerant,

highly available entity.

An instance has various parameters, which are classified in three

categories: hypervisor related-parameters (called ``hvparams``), general

parameters (called ``beparams``) and per network-card parameters (called

``nicparams``). All these parameters can be modified either at instance

level or via defaults at cluster level.

Disk template

~~~~~~~~~~~~~

The are multiple options for the storage provided to an instance; while

the instance sees the same virtual drive in all cases, the node-level

configuration varies between them.

There are five disk templates you can choose from:

diskless

  The instance has no disks. Only used for special purpose operating

  systems or for testing.

file

  The instance will use plain files as backend for its disks. No

  redundancy is provided, and this is somewhat more difficult to

  configure for high performance. Note that for security reasons the

  file storage directory must be listed under

  ``/etc/ganeti/file-storage-paths``, and that file is not copied

  automatically to all nodes by Ganeti. The format of that file is a

  newline-separated list of directories.

sharedfile

  The instance will use plain files as backend, but Ganeti assumes that

  those files will be available and in sync automatically on all nodes.

  This allows live migration and failover of instances using this

  method. As for ``file`` the file storage directory must be listed under

  ``/etc/ganeti/file-storage-paths`` or ganeti will refuse to create

  instances under it.

plain

  The instance will use LVM devices as backend for its disks. No

  redundancy is provided.

drbd

  .. note:: This is only valid for multi-node clusters using DRBD 8.0+

  A mirror is set between the local node and a remote one, which must be

  specified with the second value of the --node option. Use this option

  to obtain a highly available instance that can be failed over to a

  remote node should the primary one fail.

  .. note:: Ganeti does not support DRBD stacked devices:

     DRBD stacked setup is not fully symmetric and as such it is

     not working with live migration.

rbd

  The instance will use Volumes inside a RADOS cluster as backend for its

  disks. It will access them using the RADOS block device (RBD).

ext

  The instance will use an external storage provider. See

  :manpage:`ganeti-extstorage-interface(7)` for how to implement one.

IAllocator

~~~~~~~~~~

A framework for using external (user-provided) scripts to compute the

placement of instances on the cluster nodes. This eliminates the need to

manually specify nodes in instance add, instance moves, node evacuate,

etc.

In order for Ganeti to be able to use these scripts, they must be place

in the iallocator directory (usually ``lib/ganeti/iallocators`` under

the installation prefix, e.g. ``/usr/local``).

“Primary” and “secondary” concepts

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

An instance has a primary and depending on the disk configuration, might

also have a secondary node. The instance always runs on the primary node

and only uses its secondary node for disk replication.

Similarly, the term of primary and secondary instances when talking

about a node refers to the set of instances having the given node as

primary, respectively secondary.

Tags

~~~~

Tags are short strings that can be attached to either to cluster itself,

or to nodes or instances. They are useful as a very simplistic

information store for helping with cluster administration, for example

by attaching owner information to each instance after it's created::

  $ gnt-instance add … %instance1%

  $ gnt-instance add-tags %instance1% %owner:user2%

And then by listing each instance and its tags, this information could

be used for contacting the users of each instance.

Jobs and OpCodes

~~~~~~~~~~~~~~~~

While not directly visible by an end-user, it's useful to know that a

basic cluster operation (e.g. starting an instance) is represented

internally by Ganeti as an *OpCode* (abbreviation from operation

code). These OpCodes are executed as part of a *Job*. The OpCodes in a

single Job are processed serially by Ganeti, but different Jobs will be

processed (depending on resource availability) in parallel. They will

not be executed in the submission order, but depending on resource

availability, locks and (starting with Ganeti 2.3) priority. An earlier

job may have to wait for a lock while a newer job doesn't need any locks

and can be executed right away. Operations requiring a certain order

need to be submitted as a single job, or the client must submit one job

at a time and wait for it to finish before continuing.

For example, shutting down the entire cluster can be done by running the

command ``gnt-instance shutdown --all``, which will submit for each

instance a separate job containing the “shutdown instance” OpCode.

Prerequisites

+++++++++++++

You need to have your Ganeti cluster installed and configured before you

try any of the commands in this document. Please follow the

:doc:`install` for instructions on how to do that.

Instance management

-------------------

Adding an instance

++++++++++++++++++

The add operation might seem complex due to the many parameters it

accepts, but once you have understood the (few) required parameters and

the customisation capabilities you will see it is an easy operation.

The add operation requires at minimum five parameters:

- the OS for the instance

- the disk template

- the disk count and size

- the node specification or alternatively the iallocator to use

- and finally the instance name

The OS for the instance must be visible in the output of the command

``gnt-os list`` and specifies which guest OS to install on the instance.

The disk template specifies what kind of storage to use as backend for

the (virtual) disks presented to the instance; note that for instances

with multiple virtual disks, they all must be of the same type.

The node(s) on which the instance will run can be given either manually,

via the ``-n`` option, or computed automatically by Ganeti, if you have

installed any iallocator script.

With the above parameters in mind, the command is::

  $ gnt-instance add \

    -n %TARGET_NODE%:%SECONDARY_NODE% \

    -o %OS_TYPE% \

    -t %DISK_TEMPLATE% -s %DISK_SIZE% \

    %INSTANCE_NAME%

The instance name must be resolvable (e.g. exist in DNS) and usually

points to an address in the same subnet as the cluster itself.

The above command has the minimum required options; other options you

can give include, among others:

- The maximum/minimum memory size (``-B maxmem``, ``-B minmem``)

  (``-B memory`` can be used to specify only one size)

- The number of virtual CPUs (``-B vcpus``)

- Arguments for the NICs of the instance; by default, a single-NIC

  instance is created. The IP and/or bridge of the NIC can be changed

  via ``--net 0:ip=IP,link=BRIDGE``

See :manpage:`ganeti-instance(8)` for the detailed option list.

For example if you want to create an highly available instance, with a

single disk of 50GB and the default memory size, having primary node

``node1`` and secondary node ``node3``, use the following command::

  $ gnt-instance add -n node1:node3 -o debootstrap -t drbd -s 50G \

    instance1

There is a also a command for batch instance creation from a

specification file, see the ``batch-create`` operation in the

gnt-instance manual page.

Regular instance operations

+++++++++++++++++++++++++++

Removal

~~~~~~~

Removing an instance is even easier than creating one. This operation is

irreversible and destroys all the contents of your instance. Use with

care::

  $ gnt-instance remove %INSTANCE_NAME%

.. _instance-startup-label:

Startup/shutdown

~~~~~~~~~~~~~~~~

Instances are automatically started at instance creation time. To

manually start one which is currently stopped you can run::

  $ gnt-instance startup %INSTANCE_NAME%

Ganeti will start an instance with up to its maximum instance memory. If

not enough memory is available Ganeti will use all the available memory

down to the instance minimum memory. If not even that amount of memory

is free Ganeti will refuse to start the instance.

Note, that this will not work when an instance is in a permanently

stopped state ``offline``. In this case, you will first have to

put it back to online mode by running::

  $ gnt-instance modify --online %INSTANCE_NAME%

The command to stop the running instance is::

  $ gnt-instance shutdown %INSTANCE_NAME%

If you want to shut the instance down more permanently, so that it

does not require dynamically allocated resources (memory and vcpus),

after shutting down an instance, execute the following::

  $ gnt-instance modify --offline %INSTANCE_NAME%

.. warning:: Do not use the Xen or KVM commands directly to stop

   instances. If you run for example ``xm shutdown`` or ``xm destroy``

   on an instance Ganeti will automatically restart it (via

   the :command:`ganeti-watcher(8)` command which is launched via cron).

Querying instances

~~~~~~~~~~~~~~~~~~

There are two ways to get information about instances: listing

instances, which does a tabular output containing a given set of fields

about each instance, and querying detailed information about a set of

instances.

The command to see all the instances configured and their status is::

  $ gnt-instance list

The command can return a custom set of information when using the ``-o``

option (as always, check the manpage for a detailed specification). Each

instance will be represented on a line, thus making it easy to parse

this output via the usual shell utilities (grep, sed, etc.).

To get more detailed information about an instance, you can run::

  $ gnt-instance info %INSTANCE%

which will give a multi-line block of information about the instance,

it's hardware resources (especially its disks and their redundancy

status), etc. This is harder to parse and is more expensive than the

list operation, but returns much more detailed information.

Changing an instance's runtime memory

+++++++++++++++++++++++++++++++++++++

Ganeti will always make sure an instance has a value between its maximum

and its minimum memory available as runtime memory. As of version 2.6

Ganeti will only choose a size different than the maximum size when

starting up, failing over, or migrating an instance on a node with less

than the maximum memory available. It won't resize other instances in

order to free up space for an instance.

If you find that you need more memory on a node any instance can be

manually resized without downtime, with the command::

  $ gnt-instance modify -m %SIZE% %INSTANCE_NAME%

The same command can also be used to increase the memory available on an

instance, provided that enough free memory is available on its node, and

the specified size is not larger than the maximum memory size the

instance had when it was first booted (an instance will be unable to see

new memory above the maximum that was specified to the hypervisor at its

boot time, if it needs to grow further a reboot becomes necessary).

Export/Import

+++++++++++++

You can create a snapshot of an instance disk and its Ganeti

configuration, which then you can backup, or import into another

cluster. The way to export an instance is::

  $ gnt-backup export -n %TARGET_NODE% %INSTANCE_NAME%

The target node can be any node in the cluster with enough space under

``/srv/ganeti`` to hold the instance image. Use the ``--noshutdown``

option to snapshot an instance without rebooting it. Note that Ganeti

only keeps one snapshot for an instance - any previous snapshot of the

same instance existing cluster-wide under ``/srv/ganeti`` will be

removed by this operation: if you want to keep them, you need to move

them out of the Ganeti exports directory.

Importing an instance is similar to creating a new one, but additionally

one must specify the location of the snapshot. The command is::

  $ gnt-backup import -n %TARGET_NODE% \

    --src-node=%NODE% --src-dir=%DIR% %INSTANCE_NAME%

By default, parameters will be read from the export information, but you

can of course pass them in via the command line - most of the options

available for the command :command:`gnt-instance add` are supported here

too.

Import of foreign instances

+++++++++++++++++++++++++++

There is a possibility to import a foreign instance whose disk data is

already stored as LVM volumes without going through copying it: the disk

adoption mode.

For this, ensure that the original, non-managed instance is stopped,

then create a Ganeti instance in the usual way, except that instead of

passing the disk information you specify the current volumes::

  $ gnt-instance add -t plain -n %HOME_NODE% ... \

    --disk 0:adopt=%lv_name%[,vg=%vg_name%] %INSTANCE_NAME%

This will take over the given logical volumes, rename them to the Ganeti

standard (UUID-based), and without installing the OS on them start

directly the instance. If you configure the hypervisor similar to the

non-managed configuration that the instance had, the transition should

be seamless for the instance. For more than one disk, just pass another

disk parameter (e.g. ``--disk 1:adopt=...``).

Instance kernel selection

+++++++++++++++++++++++++

The kernel that instances uses to bootup can come either from the node,

or from instances themselves, depending on the setup.

Xen-PVM

~~~~~~~

With Xen PVM, there are three options.

First, you can use a kernel from the node, by setting the hypervisor

parameters as such:

- ``kernel_path`` to a valid file on the node (and appropriately

  ``initrd_path``)

- ``kernel_args`` optionally set to a valid Linux setting (e.g. ``ro``)

- ``root_path`` to a valid setting (e.g. ``/dev/xvda1``)

- ``bootloader_path`` and ``bootloader_args`` to empty

Alternatively, you can delegate the kernel management to instances, and

use either ``pvgrub`` or the deprecated ``pygrub``. For this, you must

install the kernels and initrds in the instance and create a valid GRUB

v1 configuration file.

For ``pvgrub`` (new in version 2.4.2), you need to set:

- ``kernel_path`` to point to the ``pvgrub`` loader present on the node

  (e.g. ``/usr/lib/xen/boot/pv-grub-x86_32.gz``)

- ``kernel_args`` to the path to the GRUB config file, relative to the

  instance (e.g. ``(hd0,0)/grub/menu.lst``)

- ``root_path`` **must** be empty

- ``bootloader_path`` and ``bootloader_args`` to empty

While ``pygrub`` is deprecated, here is how you can configure it:

- ``bootloader_path`` to the pygrub binary (e.g. ``/usr/bin/pygrub``)

- the other settings are not important

More information can be found in the Xen wiki pages for `pvgrub

<http://wiki.xensource.com/xenwiki/PvGrub>`_ and `pygrub

<http://wiki.xensource.com/xenwiki/PyGrub>`_.

KVM

~~~

For KVM also the kernel can be loaded either way.

For loading the kernels from the node, you need to set:

- ``kernel_path`` to a valid value

- ``initrd_path`` optionally set if you use an initrd

- ``kernel_args`` optionally set to a valid value (e.g. ``ro``)

If you want instead to have the instance boot from its disk (and execute

its bootloader), simply set the ``kernel_path`` parameter to an empty

string, and all the others will be ignored.

Instance HA features

--------------------

.. note:: This section only applies to multi-node clusters

.. _instance-change-primary-label:

Changing the primary node

+++++++++++++++++++++++++

There are three ways to exchange an instance's primary and secondary

nodes; the right one to choose depends on how the instance has been

created and the status of its current primary node. See

:ref:`rest-redundancy-label` for information on changing the secondary

node. Note that it's only possible to change the primary node to the

secondary and vice-versa; a direct change of the primary node with a

third node, while keeping the current secondary is not possible in a

single step, only via multiple operations as detailed in

:ref:`instance-relocation-label`.

Failing over an instance

~~~~~~~~~~~~~~~~~~~~~~~~

If an instance is built in highly available mode you can at any time

fail it over to its secondary node, even if the primary has somehow

failed and it's not up anymore. Doing it is really easy, on the master

node you can just run::

  $ gnt-instance failover %INSTANCE_NAME%

That's it. After the command completes the secondary node is now the

primary, and vice-versa.

The instance will be started with an amount of memory between its

``maxmem`` and its ``minmem`` value, depending on the free memory on its

target node, or the operation will fail if that's not possible. See

:ref:`instance-startup-label` for details.

If the instance's disk template is of type rbd, then you can specify

the target node (which can be any node) explicitly, or specify an

iallocator plugin. If you omit both, the default iallocator will be

used to determine the target node::

  $ gnt-instance failover -n %TARGET_NODE% %INSTANCE_NAME%

Live migrating an instance

~~~~~~~~~~~~~~~~~~~~~~~~~~

If an instance is built in highly available mode, it currently runs and

both its nodes are running fine, you can migrate it over to its

secondary node, without downtime. On the master node you need to run::

  $ gnt-instance migrate %INSTANCE_NAME%

The current load on the instance and its memory size will influence how

long the migration will take. In any case, for both KVM and Xen

hypervisors, the migration will be transparent to the instance.

If the destination node has less memory than the instance's current

runtime memory, but at least the instance's minimum memory available

Ganeti will automatically reduce the instance runtime memory before

migrating it, unless the ``--no-runtime-changes`` option is passed, in

which case the target node should have at least the instance's current

runtime memory free.

If the instance's disk template is of type rbd, then you can specify

the target node (which can be any node) explicitly, or specify an

iallocator plugin. If you omit both, the default iallocator will be

used to determine the target node::

   $ gnt-instance migrate -n %TARGET_NODE% %INSTANCE_NAME%

Moving an instance (offline)

~~~~~~~~~~~~~~~~~~~~~~~~~~~~

If an instance has not been create as mirrored, then the only way to

change its primary node is to execute the move command::

  $ gnt-instance move -n %NEW_NODE% %INSTANCE%

This has a few prerequisites:

- the instance must be stopped

- its current primary node must be on-line and healthy

- the disks of the instance must not have any errors

Since this operation actually copies the data from the old node to the

new node, expect it to take proportional to the size of the instance's

disks and the speed of both the nodes' I/O system and their networking.

Disk operations

+++++++++++++++

Disk failures are a common cause of errors in any server

deployment. Ganeti offers protection from single-node failure if your

instances were created in HA mode, and it also offers ways to restore

redundancy after a failure.

Preparing for disk operations

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

It is important to note that for Ganeti to be able to do any disk

operation, the Linux machines on top of which Ganeti runs must be

consistent; for LVM, this means that the LVM commands must not return

failures; it is common that after a complete disk failure, any LVM

command aborts with an error similar to::

  $ vgs

  /dev/sdb1: read failed after 0 of 4096 at 0: Input/output error

  /dev/sdb1: read failed after 0 of 4096 at 750153695232: Input/output error

  /dev/sdb1: read failed after 0 of 4096 at 0: Input/output error

  Couldn't find device with uuid 't30jmN-4Rcf-Fr5e-CURS-pawt-z0jU-m1TgeJ'.

  Couldn't find all physical volumes for volume group xenvg.

Before restoring an instance's disks to healthy status, it's needed to

fix the volume group used by Ganeti so that we can actually create and

manage the logical volumes. This is usually done in a multi-step

process:

#. first, if the disk is completely gone and LVM commands exit with

   “Couldn't find device with uuid…” then you need to run the command::

    $ vgreduce --removemissing %VOLUME_GROUP%

#. after the above command, the LVM commands should be executing

   normally (warnings are normal, but the commands will not fail

   completely).

#. if the failed disk is still visible in the output of the ``pvs``

   command, you need to deactivate it from allocations by running::

    $ pvs -x n /dev/%DISK%

At this point, the volume group should be consistent and any bad

physical volumes should not longer be available for allocation.

Note that since version 2.1 Ganeti provides some commands to automate

these two operations, see :ref:`storage-units-label`.

.. _rest-redundancy-label:

Restoring redundancy for DRBD-based instances

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

A DRBD instance has two nodes, and the storage on one of them has

failed. Depending on which node (primary or secondary) has failed, you

have three options at hand:

- if the storage on the primary node has failed, you need to re-create

  the disks on it

- if the storage on the secondary node has failed, you can either

  re-create the disks on it or change the secondary and recreate

  redundancy on the new secondary node

Of course, at any point it's possible to force re-creation of disks even

though everything is already fine.

For all three cases, the ``replace-disks`` operation can be used::

  # re-create disks on the primary node

  $ gnt-instance replace-disks -p %INSTANCE_NAME%

  # re-create disks on the current secondary

  $ gnt-instance replace-disks -s %INSTANCE_NAME%

  # change the secondary node, via manual specification

  $ gnt-instance replace-disks -n %NODE% %INSTANCE_NAME%

  # change the secondary node, via an iallocator script

  $ gnt-instance replace-disks -I %SCRIPT% %INSTANCE_NAME%

  # since Ganeti 2.1: automatically fix the primary or secondary node

  $ gnt-instance replace-disks -a %INSTANCE_NAME%

Since the process involves copying all data from the working node to the

target node, it will take a while, depending on the instance's disk

size, node I/O system and network speed. But it is (barring any network

interruption) completely transparent for the instance.

Re-creating disks for non-redundant instances

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

.. versionadded:: 2.1

For non-redundant instances, there isn't a copy (except backups) to

re-create the disks. But it's possible to at-least re-create empty

disks, after which a reinstall can be run, via the ``recreate-disks``

command::

  $ gnt-instance recreate-disks %INSTANCE%

Note that this will fail if the disks already exists. The instance can

be assigned to new nodes automatically by specifying an iallocator

through the ``--iallocator`` option.

Conversion of an instance's disk type

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

It is possible to convert between a non-redundant instance of type

``plain`` (LVM storage) and redundant ``drbd`` via the ``gnt-instance

modify`` command::

  # start with a non-redundant instance

  $ gnt-instance add -t plain ... %INSTANCE%

  # later convert it to redundant

  $ gnt-instance stop %INSTANCE%

  $ gnt-instance modify -t drbd -n %NEW_SECONDARY% %INSTANCE%

  $ gnt-instance start %INSTANCE%

  # and convert it back

  $ gnt-instance stop %INSTANCE%

  $ gnt-instance modify -t plain %INSTANCE%

  $ gnt-instance start %INSTANCE%

The conversion must be done while the instance is stopped, and

converting from plain to drbd template presents a small risk, especially

if the instance has multiple disks and/or if one node fails during the

conversion procedure). As such, it's recommended (as always) to make

sure that downtime for manual recovery is acceptable and that the

instance has up-to-date backups.

Debugging instances

+++++++++++++++++++

Accessing an instance's disks

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

From an instance's primary node you can have access to its disks. Never

ever mount the underlying logical volume manually on a fault tolerant

instance, or will break replication and your data will be

inconsistent. The correct way to access an instance's disks is to run

(on the master node, as usual) the command::

  $ gnt-instance activate-disks %INSTANCE%

And then, *on the primary node of the instance*, access the device that

gets created. For example, you could mount the given disks, then edit

files on the filesystem, etc.

Note that with partitioned disks (as opposed to whole-disk filesystems),

you will need to use a tool like :manpage:`kpartx(8)`::

  # on node1

  $ gnt-instance activate-disks %instance1%

  node3:disk/0:…

  $ ssh node3

  # on node 3

  $ kpartx -l /dev/…

  $ kpartx -a /dev/…

  $ mount /dev/mapper/… /mnt/

  # edit files under mnt as desired

  $ umount /mnt/

  $ kpartx -d /dev/…

  $ exit

  # back to node 1

After you've finished you can deactivate them with the deactivate-disks

command, which works in the same way::

  $ gnt-instance deactivate-disks %INSTANCE%

Note that if any process started by you is still using the disks, the

above command will error out, and you **must** cleanup and ensure that

the above command runs successfully before you start the instance,

otherwise the instance will suffer corruption.

Accessing an instance's console

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

The command to access a running instance's console is::

  $ gnt-instance console %INSTANCE_NAME%

Use the console normally and then type ``^]`` when done, to exit.

Other instance operations

+++++++++++++++++++++++++

Reboot

~~~~~~

There is a wrapper command for rebooting instances::

  $ gnt-instance reboot %instance2%

By default, this does the equivalent of shutting down and then starting

the instance, but it accepts parameters to perform a soft-reboot (via

the hypervisor), a hard reboot (hypervisor shutdown and then startup) or

a full one (the default, which also de-configures and then configures

again the disks of the instance).

Instance OS definitions debugging

+++++++++++++++++++++++++++++++++

Should you have any problems with instance operating systems the command

to see a complete status for all your nodes is::

   $ gnt-os diagnose

.. _instance-relocation-label:

Instance relocation

~~~~~~~~~~~~~~~~~~~

While it is not possible to move an instance from nodes ``(A, B)`` to

nodes ``(C, D)`` in a single move, it is possible to do so in a few

steps::

  # instance is located on A, B

  $ gnt-instance replace-disks -n %nodeC% %instance1%

  # instance has moved from (A, B) to (A, C)

  # we now flip the primary/secondary nodes

  $ gnt-instance migrate %instance1%

  # instance lives on (C, A)

  # we can then change A to D via:

  $ gnt-instance replace-disks -n %nodeD% %instance1%

Which brings it into the final configuration of ``(C, D)``. Note that we

needed to do two replace-disks operation (two copies of the instance

disks), because we needed to get rid of both the original nodes (A and

B).

Network Management

------------------

Ganeti used to describe NICs of an Instance with an IP, a MAC, a connectivity

link and mode. This had three major shortcomings:

  * there was no easy way to assign a unique IP to an instance

  * network info (subnet, gateway, domain, etc.) was not available on target

    node (kvm-ifup, hooks, etc)

  * one should explicitly pass L2 info (mode, and link) to every NIC

Plus there was no easy way to get the current networking overview (which

instances are on the same L2 or L3 network, which IPs are reserved, etc).

All the above required an external management tool that has an overall view

and provides the corresponding info to Ganeti.

gnt-network aims to support a big part of this functionality inside Ganeti and

abstract the network as a separate entity. Currently, a Ganeti network

provides the following:

  * A single IPv4 pool, subnet and gateway

  * Connectivity info per nodegroup (mode, link)

  * MAC prefix for each NIC inside the network

  * IPv6 prefix/Gateway related to this network

  * Tags

IP pool management ensures IP uniqueness inside this network. The user can

pass `ip=pool,network=test` and will:

1. Get the first available IP in the pool

2. Inherit the connectivity mode and link of the network's netparams

3. NIC will obtain the MAC prefix of the network

4. All network related info will be available as environment variables in

   kvm-ifup scripts and hooks, so that they can dynamically manage all

   networking-related setup on the host.

Hands on with gnt-network

+++++++++++++++++++++++++

To create a network do::

  # gnt-network add --network=192.0.2.0/24 --gateway=192.0.2.1 test

Please see all other available options (--add-reserved-ips, --mac-prefix,

--network6, --subnet6, --tags).

To make this network available on a nodegroup you should specify the

connectivity mode and link during connection::

  # gnt-network connect test bridged br100 default nodegroup1

To add a NIC inside this network::

  # gnt-instance modify --net -1:add,ip=pool,network=test inst1

This will let a NIC obtain a unique IP inside this network, and inherit the

nodegroup's netparams (bridged, br100). IP here is optional. If missing the

NIC will just get the L2 info.

To move an existing NIC from a network to another and remove its IP::

  # gnt-instance modify --net -1:ip=none,network=test1 inst1

This will release the old IP from the old IP pool and the NIC will inherit the

new nicparams.

On the above actions there is a extra option `--no-conflicts-ckeck`. This

does not check for conflicting setups. Specifically:

1. When a network is added, IPs of nodes and master are not being checked.

2. When connecting a network on a nodegroup, IPs of instances inside this

   nodegroup are not checked whether they reside inside the subnet or not.

3. When specifying explicitly a IP without passing a network, Ganeti will not

   check if this IP is included inside any available network on the nodegroup.

External components

+++++++++++++++++++

All the aforementioned steps assure NIC configuration from the Ganeti

perspective. Of course this has nothing to do, how the instance eventually will

get the desired connectivity (IPv4, IPv6, default routes, DNS info, etc) and

where will the IP resolve.  This functionality is managed by the external

components.

Let's assume that the VM will need to obtain a dynamic IP via DHCP, get a SLAAC

address, and use DHCPv6 for other configuration information (in case RFC-6106

is not supported by the client, e.g.  Windows).  This means that the following

external services are needed:

1. A DHCP server

2. An IPv6 router sending Router Advertisements

3. A DHCPv6 server exporting DNS info

4. A dynamic DNS server

These components must be configured dynamically and on a per NIC basis.

The way to do this is by using custom kvm-ifup scripts and hooks.

snf-network

~~~~~~~~~~~

The snf-network package [1,3] includes custom scripts that will provide the

aforementioned functionality. `kvm-vif-bridge` and `vif-custom` is an

alternative to `kvm-ifup` and `vif-ganeti` that take into account all network

info being exported. Their actions depend on network tags. Specifically:

`dns`: will update an external DDNS server (nsupdate on a bind server)

`ip-less-routed`: will setup routes, rules and proxy ARP

This setup assumes a pre-existing routing table along with some local

configuration and provides connectivity to instances via an external

gateway/router without requiring nodes to have an IP inside this network.

`private-filtered`: will setup ebtables rules to ensure L2 isolation on a

common bridge. Only packets with the same MAC prefix will be forwarded to the

corresponding virtual interface.

`nfdhcpd`: will update an external DHCP server

nfdhcpd

~~~~~~~

snf-network works with nfdhcpd [2,3]: a custom user space DHCP

server based on NFQUEUE. Currently, nfdhcpd replies on BOOTP/DHCP requests

originating from a tap or a bridge. Additionally in case of a routed setup it

provides a ra-stateless configuration by responding to router and neighbour

solicitations along with DHCPv6 requests for DNS options.  Its db is

dynamically updated using text files inside a local dir with inotify

(snf-network just adds a per NIC binding file with all relevant info if the

corresponding network tag is found). Still we need to mangle all these

packets and send them to the corresponding NFQUEUE.

Known shortcomings

++++++++++++++++++

Currently the following things are some know weak points of the gnt-network

design and implementation:

 * Cannot define a network without an IP pool

 * The pool defines the size of the network

 * Reserved IPs must be defined explicitly (inconvenient for a big range)

 * Cannot define an IPv6 only network

Future work

+++++++++++

Any upcoming patches should target:

 * Separate L2, L3, IPv6, IP pool info

 * Support a set of IP pools per network

 * Make IP/network in NIC object take a list of entries

 * Introduce external scripts for node configuration

   (dynamically create/destroy bridges/routes upon network connect/disconnect)