By splitting the _WaitForJobChangesHelper class into multiple smaller
classes, we gain in several places:

- Simpler code, less interaction between functions and variables
- Easy to unittest (close to 100% coverage)
- Waiting for job changes has no direct knowledge of queue anymore (it
  doesn't references queue functions anymore, especially not private ones)
- Activate inotify only if there was no change at the beginning (and
  checking again right away to avoid race conditions)

Signed-off-by: Michael Hanselmann <hansmi@google.com>
Reviewed-by: Guido Trotter <ultrotter@google.com>

* origin/devel-2.1:
  RAPI client: Implement old instance creation request format
  rlib2: Use constants for disk and NIC parameters

Conflicts:
	test/ganeti.rapi.client_unittest.py: Trivial
	test/ganeti.rapi.rlib2_unittest.py: Trivial

Signed-off-by: Michael Hanselmann <hansmi@google.com>
Reviewed-by: Iustin Pop <iustin@google.com>

Commit 8a47b447 implemented instance creation in the RAPI client,
but it left out support for the old instance creation request format.
This patch now implements the old format as good as possible. This
will only be used when talking to clusters before Ganeti 2.1.3.

Signed-off-by: Michael Hanselmann <hansmi@google.com>
Reviewed-by: Iustin Pop <iustin@google.com>

These constants were added in commit bd061c35, but the parsing code
was not updated. This also fixes a bug where a NIC's MAC address
wasn't used.

Signed-off-by: Michael Hanselmann <hansmi@google.com>
Reviewed-by: Iustin Pop <iustin@google.com>

Use RFC 5737 IP addresses and RFC 2606 domain names in all
unittests, docs, qa and docstrings.

Signed-off-by: Manuel Franceschini <livewire@google.com>
Reviewed-by: Iustin Pop <iustin@google.com>

By exposing mcpu's _Feedback function (now renamed to “Log”) to LU's,
methods like ExpandNames can also write to the job execution log.

Signed-off-by: Michael Hanselmann <hansmi@google.com>
Reviewed-by: Iustin Pop <iustin@google.com>

Since the code waiting for job changes was modified to use inotify,
a race condition between checking for changes the first time and
setting up inotify occurs. If the job is modified after the check
but before inotify is active, changes would only be noticed after
the timeout (29 seconds in most cases) expired.

Signed-off-by: Michael Hanselmann <hansmi@google.com>
Reviewed-by: Iustin Pop <iustin@google.com>

This is necessary to reuse SubmitOpCode while adding processing for
custom message types.

Signed-off-by: Michael Hanselmann <hansmi@google.com>
Reviewed-by: Iustin Pop <iustin@google.com>

Just calling utils.SafeEncode on the log message failed when it
wasn't of the type ELOG_MESSAGE and not a string. Now non-message
log entries are formatted using repr().

Signed-off-by: Michael Hanselmann <hansmi@google.com>
Reviewed-by: Iustin Pop <iustin@google.com>

The feedback function is called with only one parameter, a tuple
with the message details.

Signed-off-by: Michael Hanselmann <hansmi@google.com>
Reviewed-by: Iustin Pop <iustin@google.com>

This patch series basically adds a new parameter 'family' to the constructors
of daemon.AsyncUDPSocket and confd.client.ConfdUDPClient. This enables the
users of these two classes to support IPv6.

In ganeti-confd.ConfdAsyncUDPClient a method to check the address families of
all peers is added.

Furthermore it adds unittests for the added functionality.

Signed-off-by: Manuel Franceschini <livewire@google.com>
Reviewed-by: Iustin Pop <iustin@google.com>

Don't try to get cgroups info if instance is not running.

Signed-off-by: Balazs Lecz <leczb@google.com>
Reviewed-by: Iustin Pop <iustin@google.com>

Signed-off-by: Balazs Lecz <leczb@google.com>
Reviewed-by: Iustin Pop <iustin@google.com>

This replaces the single global log file with per-instance logs.
The instance log file is not truncated when the instance is started.

Signed-off-by: Balazs Lecz <leczb@google.com>
Reviewed-by: Iustin Pop <iustin@google.com>

* devel-2.1:
  Enable from-repository builds on old distributions

Signed-off-by: Iustin Pop <iustin@google.com>
Reviewed-by: Manuel Franceschini <livewire@google.com>

… or on distributions which simply have other implementations of man,
that do not support '--warnings'.

Signed-off-by: Iustin Pop <iustin@google.com>
Reviewed-by: Manuel Franceschini <livewire@google.com>

This patch moves network utility functions to a dedicated module.

Signed-off-by: Manuel Franceschini <livewire@google.com>
Reviewed-by: Iustin Pop <iustin@google.com>

This introduces a new instance status field, named "oper_vcpus".
It contains the actual number of VCPUs an instance is using as
seen by the hypervisor.

Signed-off-by: Balazs Lecz <leczb@google.com>
Reviewed-by: Iustin Pop <iustin@google.com>

Signed-off-by: Balazs Lecz <leczb@google.com>
Reviewed-by: Iustin Pop <iustin@google.com>

Signed-off-by: Apollon Oikonomopoulos <apollon@noc.grnet.gr>
Signed-off-by: Iustin Pop <iustin@google.com>
Reviewed-by: Iustin Pop <iustin@google.com>

LUCreateInstance uses the cluster-wide default iallocator if no iallocator or
primary node is specified manually.

Signed-off-by: Apollon Oikonomopoulos <apollon@noc.grnet.gr>
Signed-off-by: Iustin Pop <iustin@google.com>
Reviewed-by: Iustin Pop <iustin@google.com>

Add unit tests to check the function of _CheckIAllocatorOrNode

Signed-off-by: Apollon Oikonomopoulos <apollon@noc.grnet.gr>
Signed-off-by: Iustin Pop <iustin@google.com>
Reviewed-by: Iustin Pop <iustin@google.com>

_CheckIAllocatorOrNode will be called by LUs wishing to use an instance
allocator or a target node. It performs sanity checks and will modify the LU's
opcode's iallocator slot to use the cluster-wide allocator if
appropriate.

Signed-off-by: Apollon Oikonomopoulos <apollon@noc.grnet.gr>
Signed-off-by: Iustin Pop <iustin@google.com>
Reviewed-by: Iustin Pop <iustin@google.com>

Signed-off-by: Apollon Oikonomopoulos <apollon@noc.grnet.gr>
Signed-off-by: Iustin Pop <iustin@google.com>
Reviewed-by: Iustin Pop <iustin@google.com>

Add a cluster parameter to hold the iallocator that will be used by default
when required and no alternative (manually-specified iallocator or
manually-specified node(s)) is given.

Signed-off-by: Apollon Oikonomopoulos <apollon@noc.grnet.gr>
Signed-off-by: Iustin Pop <iustin@google.com>
Reviewed-by: Iustin Pop <iustin@google.com>

Signed-off-by: Balazs Lecz <leczb@google.com>
Reviewed-by: Iustin Pop <iustin@google.com>

Signed-off-by: Luca Bigliardi <shammash@google.com>
Reviewed-by: Iustin Pop <iustin@google.com>

Node daemon prints a lot of warnings if --no-mlock option is not specified and
ctypes module is not present.

With the following patch the warning is printed only at noded startup.

Signed-off-by: Luca Bigliardi <shammash@google.com>
Reviewed-by: Michael Hanselmann <hansmi@google.com>

I have seen some very seldom errors where (it seems) the address is
still live for a short while after removing it from the old master, thus
the new master will fail in startup/adding its own IP address.

To prevent against this, we add a delay/retry before we proceed, if the
IP is still reachable.

Signed-off-by: Iustin Pop <iustin@google.com>
Reviewed-by: René Nussbaumer <rn@google.com>
(cherry picked from commit 425f0f54)

Signed-off-by: Iustin Pop <iustin@google.com>
Reviewed-by: Balazs Lecz <leczb@google.com>

Signed-off-by: Balazs Lecz <leczb@google.com>
Reviewed-by: Iustin Pop <iustin@google.com>

Signed-off-by: Balazs Lecz <leczb@google.com>
Reviewed-by: Iustin Pop <iustin@google.com>

* devel-2.1:
  QA, burnin: allow selection of reboot types
  Add a QA option to disable reboots during burnin

Conflicts:
	qa/qa-sample.json (trivial)
	qa/qa_cluster.py  (trivial)

Signed-off-by: Iustin Pop <iustin@google.com>
Reviewed-by: Luca Bigliardi <shammash@google.com>

After some more investigation, only the soft reboot type fails for Xen
3.4 (due to the reboot/uptime time counter). As such, it's better to
allow selective testing, since we do want to test in general these
opcodes/the command line script.

Signed-off-by: Iustin Pop <iustin@google.com>
Reviewed-by: Luca Bigliardi <shammash@google.com>

Signed-off-by: Iustin Pop <iustin@google.com>
Reviewed-by: Michael Hanselmann <hansmi@google.com>

Currently, the way to signal export failures is by the return value.
This means that if a client doesn't check the values (e.g. burnin), any
failure is being ignore. And this is what we've been doing forever in
burning (not actually testing that the export is successful).

This patch changes the behaviour of ExportInstance: it will abort with
an exception for any error, and removes the custom handling from
gnt-backup. This makes the behaviour consistent for any client (e.g.
RAPI), and it prevents false positives. If, for a given instance, a
subset of disks should not be backed up, the OS scripts should handle
that case.

Signed-off-by: Iustin Pop <iustin@google.com>
Reviewed-by: Michael Hanselmann <hansmi@google.com>

Currently, GenericManyOps ignores the actual success or failure results
from the invididual jobs. We change this to return '0' (i.e. success)
only when all jobs failed, as many times we have just one job.

Together with the JobExecutor change, this will report failures
correctly when used with a drained queue and submit only, or when used
normally and the opcode actually fails.

Signed-off-by: Iustin Pop <iustin@google.com>
Reviewed-by: Michael Hanselmann <hansmi@google.com>

Currently, for the 'wait' case, we return a list of tuples (status,
result), in the order of submitted jobs, but we don't return anything
for the no-wait case.

This patch changes the no-wait case to return a list of tuples (status,
result), where result can be either a job ID or an error message.
Processing in clients can then ignore whether we did wait or not, and
test the overall or individual status.

Signed-off-by: Iustin Pop <iustin@google.com>
Reviewed-by: Michael Hanselmann <hansmi@google.com>

With the recent changes in the job queue, an old bug surfaced: we never
serialized the status change when in NotifyStart, thus a crash of the
master would have left the job queue oblivious to the fact that the job
was actually running.

In the previous implementation, queries against the job status were
using the in-memory object, so they 'saw' and reported correctly the
running status. But the new implementation just looks at the on-disk
version, and thus didn't see this transition.

The patch also moves NotifyStart to a decorator-based version (like the
other functions), which generates a lot of churn in the diff, sorry.

Signed-off-by: Iustin Pop <iustin@google.com>
Reviewed-by: Michael Hanselmann <hansmi@google.com>

Signed-off-by: Balazs Lecz <leczb@google.com>
Reviewed-by: Iustin Pop <iustin@google.com>

Signed-off-by: Balazs Lecz <leczb@google.com>
Reviewed-by: Iustin Pop <iustin@google.com>