GitLab

This patch modifies how the KVM daemon is started/stopped.  After this
patch, the KVM daemon is always started by default, as it will
terminate itself if necessary, and it is restarted by the watcher.

Also, changes to user shutdown and vm capable attributes will cause
the KVM daemon instances on the corresponding nodes to be
started/stopped accordingly.  This is achieved through the helper
function 'EnsureKvmdOnNodes'.
Signed-off-by: Jose A. Lopes <jabolopes@google.com>
Reviewed-by: Hrvoje Ribicic <riba@google.com>

This RPC will be needed so that changes to the cluster parameters for
enabled hypervisors, user shutdown, and vm capable nodes, can
start/stop the KVM daemon as necessary.
Signed-off-by: Jose A. Lopes <jabolopes@google.com>
Reviewed-by: Hrvoje Ribicic <riba@google.com>
Reviewed-by: Helga Velroyen <helgav@google.com>

This patch modifies the KVM daemon startup to test whether the node if
the enabled hypervisor is KVM, whether the node is VM capable, and if
user shutdown is enabled.  If at least one of these conditions is not
met, the KVM daemon will exit.
Signed-off-by: Jose A. Lopes <jabolopes@google.com>
Reviewed-by: Hrvoje Ribicic <riba@google.com>

* Add list of enabled hypervisors, cluster wide parameter for user
  shutdown, and a list of nodes and their respective vm capable
  attributes to Ssconf.  This will be needed by the KVM daemon to
  determine whether it should start on a particular node.

* Extend unit tests
Signed-off-by: Jose A. Lopes <jabolopes@google.com>
Reviewed-by: Hrvoje Ribicic <riba@google.com>
Reviewed-by: Helga Velroyen <helgav@google.com>

... including object, queries, opcode, LU, command line, upgrade, etc.
Signed-off-by: Jose A. Lopes <jabolopes@google.com>
Reviewed-by: Hrvoje Ribicic <riba@google.com>
Reviewed-by: Helga Velroyen <helgav@google.com>

Signed-off-by: Jose A. Lopes <jabolopes@google.com>
Reviewed-by: Hrvoje Ribicic <riba@google.com>

... to reduce code duplication.
Signed-off-by: Jose A. Lopes <jabolopes@google.com>
Reviewed-by: Hrvoje Ribicic <riba@google.com>

Signed-off-by: Jose A. Lopes <jabolopes@google.com>
Reviewed-by: Hrvoje Ribicic <riba@google.com>

... in 'seqSplit' and 'parseChoices'.
Signed-off-by: Jose A. Lopes <jabolopes@google.com>
Reviewed-by: Hrvoje Ribicic <riba@google.com>

... in the KVM hypervisor module.
Signed-off-by: Jose A. Lopes <jabolopes@google.com>
Reviewed-by: Hrvoje Ribicic <riba@google.com>

Signed-off-by: Jose A. Lopes <jabolopes@google.com>
Reviewed-by: Hrvoje Ribicic <riba@google.com>

Signed-off-by: Jose A. Lopes <jabolopes@google.com>
Reviewed-by: Helga Velroyen <helgav@google.com>

The user shutdown parameter was being enabled for the instance before
the test starts but afterwards it was not being disabled.
Signed-off-by: Jose A. Lopes <jabolopes@google.com>
Reviewed-by: Reviewed-by: Hrvoje Ribicic <riba@google.com>

* stable-2.10
  (no changes)

* stable-2.9
  (no changes)

* stable-2.8:
  Fix QA patching functionality
Signed-off-by: Hrvoje Ribicic <riba@google.com>
Reviewed-by: Petr Pudlak <pudlak@google.com>

* stable-2.9
  (no changes)

* stable-2.8:
  Fix QA patching functionality
Signed-off-by: Hrvoje Ribicic <riba@google.com>
Reviewed-by: Petr Pudlak <pudlak@google.com>

* stable-2.8:
  Fix QA patching functionality
Signed-off-by: Hrvoje Ribicic <riba@google.com>
Reviewed-by: Petr Pudlak <pudlak@google.com>

This patch clarifies the code a bit and fixes a bug preventing QA
patches from working.
Signed-off-by: Hrvoje Ribicic <riba@google.com>
Reviewed-by: Klaus Aehlig <aehlig@google.com>

* stable-2.10
  Revision bump for 2.10.5
  Add NEWS file for 2.10.5
  Check drbd helper only in VM capable nodes
  'Raise' called inside 'CheckPrereq' needs the prereq kw

Conflicts:
	NEWS: take all news entries
	configure.ac: ignore revision bump
Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Petr Pudlak <pudlak@google.com>

Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Jose Lopes <jabolopes@google.com>

This version is to be released today. The major changes are
the modified htools allocation computation and the addition
of two new options to 'gnt-group evacuate'.
Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Jose Lopes <jabolopes@google.com>

Before having the watcher cleanup down instances, the assumption was
that if the instance data was contained in the watcher's state, then
the number of restart attempts would definitely be present.  This is
not true anymore, because it might be the case that if the instance
data is in the watcher's state, it is possible that only one of the
keys is present (or both).  This patch fixes this assumption.
Signed-off-by: Jose A. Lopes <jabolopes@google.com>
Reviewed-by: Hrvoje Ribicic <riba@google.com>

Modify 'Hypervisor.ListInstances' to exclude instances that have been
shutdown by the user.  In other words, 'ListInstances' includes only
instances that are actually running.  This modification requires also
modifying the functions in the 'ganeti.backend' module that start,
stop, and reboot an instance, in order to work properly with user
shutdown.
Signed-off-by: Jose A. Lopes <jabolopes@google.com>
Reviewed-by: Hrvoje Ribicic <riba@google.com>

Extend QA for instance user down with assertions for 'admin_state' and
'admin_state_source', and modify test to try different operation order
for instance user down, stop, start, etc.
Signed-off-by: Jose A. Lopes <jabolopes@google.com>
Reviewed-by: Hrvoje Ribicic <riba@google.com>

* Given the previous changes to user down report status, we can lift
  the restriction that user down instances could not be directly
  started.  This patch modifies instance startup to first cleanup user
  down instances before starting them.

* Also, the opcode for instance create is given a shutdown timeout and
  the unit tests are updated accordingly.
Signed-off-by: Jose A. Lopes <jabolopes@google.com>
Reviewed-by: Hrvoje Ribicic <riba@google.com>

* Query addtional instance fields, namely, 'admin_state' and
  'admin_state_source'.  Together with the instance field 'status',
  the watcher is able to determine whether an instance was shutdown by
  the user and whether it was already cleaned up.

* Add 'Instance' method 'NeedsCleanup' which determines whether a user
  down instance needs to be cleaned up (i.e., it was not already
  cleaned up).
Signed-off-by: Jose A. Lopes <jabolopes@google.com>
Reviewed-by: Hrvoje Ribicic <riba@google.com>

Modify opcode and LU for instance shutdown to mark an instance as user
down.  This is used by the watcher to cleanup instances that were
shutdown by the user while retaining the instance status 'USER_down'.
Signed-off-by: Jose A. Lopes <jabolopes@google.com>
Reviewed-by: Hrvoje Ribicic <riba@google.com>

* Fix instance queries (gnt-instance list and info) in Python and
  Haskell to report 'USER_down' when the instance is not running but
  the admin state is 'ADMINST_down' and the admin state source is the
  user.

* Fix 'hvparams' to get the dict filled with the cluster defaults.
Signed-off-by: Jose A. Lopes <jabolopes@google.com>
Reviewed-by: Hrvoje Ribicic <riba@google.com>

Extend the configuration to update the instance field
'admin_state_source' whenever there is a change to the instance field
'admin_state'.
Signed-off-by: Jose A. Lopes <jabolopes@google.com>
Reviewed-by: Hrvoje Ribicic <riba@google.com>

Add 'Instance' field 'admin_state_source' to data of unit tests.
Signed-off-by: Jose A. Lopes <jabolopes@google.com>
Reviewed-by: Hrvoje Ribicic <riba@google.com>

... which keeps track of who last changed the instance field
'admin_state', namely, the admin or the user.
Signed-off-by: Jose A. Lopes <jabolopes@google.com>
Reviewed-by: Hrvoje Ribicic <riba@google.com>

* Add type 'AdminStateSource' which keeps track of who last changed
  the instance field 'admin_state': the admin or the user.

* Add constants 'adminSource', 'userSource', and 'adminStateSources',
  which are needed by the instance shutdown opcode and LU in Python.

* Add type predicate 'TAdminStateSource', which is needed by the
  opcode parameter.
Signed-off-by: Jose A. Lopes <jabolopes@google.com>
Reviewed-by: Hrvoje Ribicic <riba@google.com>

Luxid would always show user-shutdown instances as ADMIN_down when user
shutdown was not allowed, because it was checking for only two
possibilities:

  - Either the instance is admin-up and user-down (allowed), or
  - The instance is admin-down

However there is a third state: the instance being admin-up but not
running (or user-down but disallowed). We fix this by checking for the
additional state.
Signed-off-by: Apollon Oikonomopoulos <apoikos@gmail.com>
Reviewed-by: Jose A. Lopes <jabolopes@google.com>

Info queries on user-down instances would fail, because of a check
against instance.hvparams which only contains the instance's non-default
parameters. To fix this, we use GetUpdatedParams, to merge with the cluster
hypervisor parameters.

We also fix a logic error uncovered after the above fix: instances that
were shut down by the user but not marked for user_shutdown, would be
reported as "up" instead of "down".
Signed-off-by: Apollon Oikonomopoulos <apoikos@gmail.com>
Reviewed-by: Jose A. Lopes <jabolopes@google.com>

... instead of using the RPC fail message directly.
Signed-off-by: Jose A. Lopes <jabolopes@google.com>
Reviewed-by: Hrvoje Ribicic <riba@google.com>

.. and additional local variables
Signed-off-by: Jose A. Lopes <jabolopes@google.com>
Reviewed-by: Hrvoje Ribicic <riba@google.com>

... as per the definition of 'ListInstances' and Python convention.
Signed-off-by: Jose A. Lopes <jabolopes@google.com>
Reviewed-by: Hrvoje Ribicic <riba@google.com>

Signed-off-by: Jose A. Lopes <jabolopes@google.com>
Reviewed-by: Hrvoje Ribicic <riba@google.com>

Signed-off-by: Jose A. Lopes <jabolopes@google.com>
Reviewed-by: Hrvoje Ribicic <riba@google.com>

Upon change of enabled disk templates, Ganeti checks for a valid drbd
helper among all nodes. Let those nodes be only the VM capable ones.
Signed-off-by: Dimitris Aragiorgis <dimara@grnet.gr>
Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Klaus Aehlig <aehlig@google.com>

This patch fixes the missing 'prereq' keyword in calls to 'Raise' in
the control flow of 'CheckPrereq', and updates the tests.
Signed-off-by: Jose A. Lopes <jabolopes@google.com>
Reviewed-by: Hrvoje Ribicic <riba@google.com>