GitLab

As in the previous patch, the option '--no-drbd-storage'
is deprectated, because it is subsumed by the non-inclusion
of 'drbd' in the list of enabled disk templates.
Signed-off-by: Helga Velroyen <helgav@google.com>
Reviewed-by: Thomas Thrainer <thomasth@google.com>

Whether or not a particular type of storage is enabled
or not is determined by the list of enabled disk templates
in the cluster's configuration. This makes the option
'--no-drbd-storage' obsolete, because it is subsumed by
not including 'drbd' in the list of enabled disk templates.
Signed-off-by: Helga Velroyen <helgav@google.com>
Reviewed-by: Thomas Thrainer <thomasth@google.com>

This change makes sure that 'gnt-cluster verify' only
checks for the DRBD usermode helper, if DRBD is actually
enabled.
Signed-off-by: Helga Velroyen <helgav@google.com>
Reviewed-by: Thomas Thrainer <thomasth@google.com>

This patch factors out the functions that deal with setting
and modifying the DRBD usermode helper in cluster.py in
order to make them more unittestable. The unit tests are
provided as well. No functional changes otherwise.
Signed-off-by: Helga Velroyen <helgav@google.com>
Reviewed-by: Thomas Thrainer <thomasth@google.com>

This patch factors out the functions in gnt_cluster
(related to cluster init and cluster modify) which
deal with setting / determining the drbd usermode helper
to make them more testable. Unit tests for the extracted
functions are provided as well. No function changes
otherwise.
Signed-off-by: Helga Velroyen <helgav@google.com>
Reviewed-by: Thomas Thrainer <thomasth@google.com>

This patch factors out the function that checks the DRBD
usermode helper in bootstrap (cluster init) in order to make
it more testable. It also contains the unit tests. Otherwise,
no functional changes.
Signed-off-by: Helga Velroyen <helgav@google.com>
Reviewed-by: Thomas Thrainer <thomasth@google.com>

Ganeti now installs all its files into a version-specific directory
and only adds symbolic links at the canonical installation places.
Nevertheless, make documentation, in particular man pages, still contain
the old places, i.e., the places where the symbolic links are added.
Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Guido Trotter <ultrotter@google.com>

Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Guido Trotter <ultrotter@google.com>

Even though the ${PREFIX}/opt/ only came into the design in the review
process, it is not the best choice, as /opt is reserverd for manually
installed software, but Ganeti intends to be packaged. Use ${libdir}
instead, defaulting to ${PREFIX}/lib.
Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Guido Trotter <ultrotter@google.com>

Function '_CheckOutputFields' was receiving three field sets, two of
which were simply being merged together.  This was complicated and
unnecessary.  This patch simplifies '_CheckOutputFields' to take
simply two field sets.
Signed-off-by: Jose A. Lopes <jabolopes@google.com>
Reviewed-by: Thomas Thrainer <thomasth@google.com>

Constant 'SF_NODE' is a storage field and, therefore, should be in
'VALID_STORAGE_FIELDS'.  This patch fixes this and also reference to
these constants, namely, in 'cmdlib' and 'qa'.
Signed-off-by: Jose A. Lopes <jabolopes@google.com>
Reviewed-by: Thomas Thrainer <thomasth@google.com>

QA was adding 'SF_TYPE' to a list containing all storage fields (i.e.,
'VALID_STORAGE_FIELDS').  However, 'SF_TYPE' is already part of
'VALID_STORAGE_FIELDS', which means this element was occurring twice
in the list.
Signed-off-by: Jose A. Lopes <jabolopes@google.com>
Reviewed-by: Thomas Thrainer <thomasth@google.com>

Add "MonD data" sub-section.
Signed-off-by: Spyros Trigazis <strigazi@gmail.com>
Signed-off-by: Michele Tartara <mtartara@google.com>
Reviewed-by: Michele Tartara <mtartara@google.com>

That was forgotten when the parameter was added.
Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Michele Tartara <mtartara@google.com>

In commit 807d8853

, new node parameters were introduced in the python world.
Add them to the haskell world as well to restore consistency.
Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Michele Tartara <mtartara@google.com>

As of Ganeti 2.10, all files are installed into a single, version
specific, directory. In this way, several Ganeti versions can be
installed in parallel. Appropriate symbolic links in the search path
will be added by the Ganeti upgrade procedure. For the initial
installation, however, links in the search path are needed. So make
the installation of symbolic links a configure-time option, defaulting
to the current behavior, that that installation is visible in the
path.
Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Thomas Thrainer <thomasth@google.com>

As first step of implementing the new upgrade procedure (see design
document), make Ganeti install all files into a single directory,
$(prefix)/opt/Ganeti-$(VERSION). Symbolic links outside to this directory
will be added (either at install time, or by a version specific install
script at upgrade/downgrade time). This will make it possible to have
multiple version of Ganeti installed at the same machine, with only one
being active.
Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Thomas Thrainer <thomasth@google.com>

Fix two grammatical errors in gnt-node man page
Signed-off-by: Sebastian Gebhard <sege@fs.ei.tum.de>
Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Klaus Aehlig <aehlig@google.com>

This will add the new node parameters to the ganeti man page.
Signed-off-by: Sebastian Gebhard <sege@fs.ei.tum.de>
Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Klaus Aehlig <aehlig@google.com>

Extend gnt-node by parameters for OpenvSwitch. --ovs Flag to enable
OpenvSwitch, --ovs-name for the name of the OpenvSwitch and --ovs-link
for the interface to connect to.
Signed-off-by: Sebastian Gebhard <sege@fs.ei.tum.de>
Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Klaus Aehlig <aehlig@google.com>

This patches adds the needed RPC functions in order to create the
OpenvSwitches on the nodes.
Signed-off-by: Sebastian Gebhard <sege@fs.ei.tum.de>
Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Klaus Aehlig <aehlig@google.com>

This patch adds functionality to LUNodeAdd to
  - check the arguments given. It will warn if no physical link is given
    and fail if OpenvSwitch is not enabled, but parameters are given
  - call the RPC to configure OpenvSwitch on the node
Signed-off-by: Sebastian Gebhard <sege@fs.ei.tum.de>
Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Klaus Aehlig <aehlig@google.com>

This is the functionality to create the OpenvSwitches on the nodes.
Parameters are given via opcode and checked as well as extended with
the defaults.
Signed-off-by: Sebastian Gebhard <sege@fs.ei.tum.de>
Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Klaus Aehlig <aehlig@google.com>

This patch introduces the needed fields for OpenvSwitch parameters
into ndparams and also provides the default values.

The parameters are:
ND_OVS: boolean, to show whether OpenvSwitch is enabled or not
ND_OVS_NAME: the name of the OpenvSwitch to create
ND_OVS_LINK: the physical link to the OpenvSwitch

While ND_OVS and ND_OVS_NAME are mandatory and default to FALSE and
constants.DEFAULT_OVS respectively, ND_OVS_LINK is optional and can
be left empty.
Adding a new node with OpenvSwitch and no physical interface might
or might not be what the user wants, so this case will result in a
warning to inform the user (see later patch in this series).

This patch also fixes unittests which are using these parameters.
Signed-off-by: Sebastian Gebhard <sege@fs.ei.tum.de>
Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Klaus Aehlig <aehlig@google.com>

This design doc adds details about how the autoconfiguration of
openvswitch should work in the cluster <-> node group <-> node
hierarchy. Parameters that are needed for succesful configration
are idenfied and inheritance of parameters is defined.

Also, some rephrasing of the definition in the first part of the
document. After looking into gnt-network, in my opinion it is not the
right place to put the switch management into, since it has more to
do with links and nics than with network names and addresses.
Signed-off-by: Sebastian Gebhard <sege@fs.ei.tum.de>
Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Klaus Aehlig <aehlig@google.com>

The enabled disk templates in IPolicies are stricter checked after the
merge from 2.9, so adapt the tests to follow those changes.
Signed-off-by: Thomas Thrainer <thomasth@google.com>
Reviewed-by: Helga Velroyen <helgav@google.com>

This patch adds unit test coverage for LUInstanceMultiAlloc.
Signed-off-by: Thomas Thrainer <thomasth@google.com>
Reviewed-by: Guido Trotter <ultrotter@google.com>

This patch adds unit test coverage for LUInstanceRename.
Signed-off-by: Thomas Thrainer <thomasth@google.com>
Reviewed-by: Guido Trotter <ultrotter@google.com>

This patch adds basic unit tests for instance import in
LUInstanceCreate. It does not provide full coverage though.
Signed-off-by: Thomas Thrainer <thomasth@google.com>
Reviewed-by: Guido Trotter <ultrotter@google.com>

Introduce a new method of creating a mocked LU for tests, and reduce the
use of the legacy _FakeLU class.
Signed-off-by: Thomas Thrainer <thomasth@google.com>
Reviewed-by: Guido Trotter <ultrotter@google.com>

This patch adds unit test coverage for LUInstanceRemove and
LUInstanceMove.
Signed-off-by: Thomas Thrainer <thomasth@google.com>
Reviewed-by: Guido Trotter <ultrotter@google.com>

- Move instance related tests to instance_unittest.py
- Adapt moved tests to the new test framework where appropriate
- Add unit test coverage for LUInstanceCreate
  - Only instance creation is covered yet, no imports
Signed-off-by: Thomas Thrainer <thomasth@google.com>
Reviewed-by: Guido Trotter <ultrotter@google.com>

The disk_template parameter is optional during disk create (the first
enabled disk template is taken in this case), so don't require it.
Signed-off-by: Thomas Thrainer <thomasth@google.com>
Reviewed-by: Guido Trotter <ultrotter@google.com>

* stable-2.9
  Allow instance mods if only osparams are given
  Use node UUIDs for locking instead of node names
  Allow instance mods if only osparams are given
  Fix a node name vs. UUID bug in instance import
  Typo in hroller man page
  Use node UUID for locking in LUInstanceMove
  Fix harep manpage title
  Fix IPolicy violation check in LUGroupSetParams
  Use FQDN to check master node status
  Revert "Display node name instead of UUID in error message"
  Display node name instead of UUID in error message
  Use node name in error message
  QA: make ipolicy test respect enabled disk templates
  QA: adjust tests wrt to ipolicy disk templates
  ClusterSetParams: move vg-name checks from to CheckPrereq
  man gnt-cluster: mention ipolicy check
  Update NEWS file regarding ipolicy checks
  gnt-group add/modify: ipolicy vs disk templates
  Move Ipolicy utility function to cmdlib/common.py
  bootstrap: restrict ipolicy to enabled disk templates
  gnt-cluster modify: ipolicy vs enabled disk templates
  gnt-cluster modify: factor out ipolicy check

Conflicts:
    lib/cmdlib/group.py (due to moved function, changes already in
                         master)
Signed-off-by: Thomas Thrainer <thomasth@google.com>
Reviewed-by: Michele Tartara <mtartara@google.com>

This design document describes how upgrade and downgrades
can be made more automatic for future versions.
Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Michele Tartara <mtartara@google.com>

* stable-2.8
  Allow instance mods if only osparams are given
  Fix harep manpage title
  Use FQDN to check master node status
Signed-off-by: Thomas Thrainer <thomasth@google.com>
Reviewed-by: Klaus Aehlig <aehlig@google.com>

osparams are now recognized as changes, so changing only those without
other changes works too now.

(cherry picked from commit 5eae613c

)
Signed-off-by: Thomas Thrainer <thomasth@google.com>
Reviewed-by: Helga Velroyen <helgav@google.com>
Reviewed-by: Guido Trotter <ultrotter@google.com>

LUInstanceMultiAlloc used node names instead of node UUIDs to lock
required nodes. This patch fixes this bug.
Signed-off-by: Thomas Thrainer <thomasth@google.com>
Reviewed-by: Guido Trotter <ultrotter@google.com>

Running 'make' in the chroot complains about pcre 0.94.4
being required.
Signed-off-by: Helga Velroyen <helgav@google.com>
Reviewed-by: Michele Tartara <mtartara@google.com>

osparams are now recognized as changes, so changing only those without
other changes works too now.
Signed-off-by: Thomas Thrainer <thomasth@google.com>
Reviewed-by: Helga Velroyen <helgav@google.com>