* devel-2.1:
  Add disk cache control parameter for KVM
  Change pyinotify import for broader compatibility
  ClusterMasterQuery: add primary ip field
  confd ClusterMasterQuery: allow fields request
  Simplify utils.ReadFile
  DRBD: ignore unreadable meta devices
  gnt-cluster verify: Warn if node time diverges too far
  KVM: fail when a routed nic has no ip
  Enable batch mode for devel/upload
  cmdlib: Work around race condition in DRBD before version 8.0.13

This patch adds the 'cache' parameter for KVM; currently this is only
customisable at the hypervisor level, so it's the same for all drives
(except any CDROM image, which gets the default).

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

On some distributions pyinotify is installed in a different way, and the
actual module just contains an internal pyinotify entry, which is the
actual library. On others the main pyinotify module contains the library
itself. We'll try both, in order to be more compatible.

Signed-off-by: Guido Trotter <ultrotter@google.com>

By allowing also the primary ip field to be fetched directly, we avoid
one more confd lookup, or dns request, to find out which address the
master node lives at.

Signed-off-by: Guido Trotter <ultrotter@google.com>

Change the ClusterMasterQuery to allow a query, and if present accept a
list of fields to be returned. Currently only name and ip are accepted.

This feature will be used by NLD to route the cluster ip over the nbma.

Backwards compatibility is preserved.

Signed-off-by: Guido Trotter <ultrotter@google.com>

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

Pass SSL certificate and key paths from ganeti-* instead of using a dict. The
patch also improves the --ssl-{key,cert} options by giving the default in
--help output and changes the validation a bit.

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

Passing it in as a parameter seems more logical.

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

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

The documentation for file objects' read method states that if the size
is "negative or ommitted", all data is read; thus we can simplify it to
have size=-1 by default and not have the if test.

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

The DRBD driver can ignore dead disks but not dead meta devices (for
which it refuses to configure the minor). To handle this case, we check
that the meta device is readable and if not we ignore it (the same as
when backend._RecursiveAssembleBD didn't find it).

A needed change is the move of some checks and initialisers before this
test (which is before the super().__init__ call), but that should be
fine.

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

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

The warning will be generated if the clocks diverge by more
than 150 seconds. Due to the way the RPC system works, we
cannot get exact time differences, e.g. if one of the
queried nodes is broken. The comparision is done using a
time window.

Confd queries will fail if the clock on the client and server
are more than 300 seconds from each other. This check helps
keeping at least the nodes of a cluster in sync.

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

Signed-off-by: Guido Trotter <ultrotter@google.com>

This shouldn't happen, but if it does it's better to fail at this level,
rather than create a broken NIC script, which is hard to debug.

Signed-off-by: Guido Trotter <ultrotter@google.com>
Reviewed-by: Iustin Pop <iustin@google.com>

Since the rsync/ssh calls are done in parallel, they can't read properly a
password or confirmation about keys from stdin. As such, it's better to enable
batch mode so that they fail right away instead of prompting and then timing
out after a long while.

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

DRBD goes into sync mode for a short amount of time after
executing the "resize" command. DRBD 8.x below version
8.0.13 contains a bug whereby calling "resize" in sync
mode fails.

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

This patch adds the 'cache' parameter for KVM; currently this is only
customisable at the hypervisor level, so it's the same for all drives
(except any CDROM image, which gets the default).

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

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

It can't work on older distributions where simplejson
doesn't have indentation support.

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

When not initialising the cluster, consider all nodes are added, so that
multi-node tests (e.g. export, replace) work correctly (if there are
nodes, of course).

Signed-off-by: Iustin Pop <iustin@google.com>
Reviewed-by: Guido Trotter <ultrotter@google.com>

This patch removes the quotes from CommaJoin and converts most of the
callers (that I could find) to it. Since CommaJoin does str(i) for i in
param, we can remove these, thus simplifying slightly a few calls.

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

Commit 495cfdf0 removed “nic.bridges” from the default
list for bulk instance list RAPI requests.

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

Comparing a number with None is not a good idea:

  >>> (0 < None, 0 > None)
  (False, True)

This patch also adds build-bash-completion to the list
of checked Python scripts and wraps one line of more
than 80 characters.

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

This reverts commit 6915bc28 based on thread on
ganeti-devel.

Conflicts:

	lib/cmdlib.py (due to the error code classification, trivial)

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

By checking for this explicitly, the errors (SSLEAY_RAND_BYTES, “PRNG
not seeded”) will happen in the start-up phase of the daemon and not
only when executing remote procedure calls.

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

Clarify the fact that temporary HV/BE params in instance start override
and do not extend the configured parameters; and change the instance
list headers from HVM_* to * since many of the parameters apply to KVM
too. Also fix a typo in the rapi documention for '/2/nodes'.

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

This should fix an issue I've seen exactly once during testing. It might have
been caused by parallel RPC calls to archive jobs.

[…] ganeti-noded:112 ERROR Error in RPC call […]
 File "/usr/lib/python2.4/site-packages/ganeti/backend.py", line 2365, in JobQueueRename
   utils.RenameFile(old, new, mkdir=True)
 File "/usr/lib/python2.4/site-packages/ganeti/utils.py", line 322, in RenameFile
   os.makedirs(os.path.dirname(new), mkdir_mode)
 File "/usr/lib/python2.4/os.py", line 159, in makedirs
   mkdir(name, mode)
OSError: [Errno 17] File exists: '/var/lib/ganeti/queue/archive/0'

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

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

For an instance with exactly one NIC:

$ gnt-instance modify --net 1:ip=1.2.3.4 inst1
Failure: prerequisites not met for this operation:
error type: wrong_input, error details:
Invalid NIC index 1, valid values are 0 to 1

For an instance with no NIC at all, it fails with “Invalid NIC index 0, valid
values are 0 to 0”. This is fixed by this patch.

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

This was originally implemented in 0ce8f948 and partially
rolled back in 9b65e0d4. Apart from re-adding the check,
this patch does some housekeeping by renaming the “_helper”
function to “_AddIpAddress”.

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

(1) Both the man page and the online help report the link and mode
fields, which are in the code called nic_link and nic_mode.
(2) Add missing fields to the online help.

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

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

This includes a few paragraphs about daemon-util.

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

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

This makes it easier to catch warnings.

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

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

This is in preparation to including it into the large
documentation.

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

To stay on the safe side, we check for errors in all instances, and
refuse to act, reporting on the errors we found, if there are any
problems.

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

We should match for the strings to be the same "==" not to point to the
same memory location with is, or we skip the actual check.

Signed-off-by: Guido Trotter <ultrotter@google.com>