While git-am'ing the blockdev_prefix patch, I modified the hv_xen.py
file but forgot to amend the patch :(, so here it is.

Furthermore, to detect such inconsistencies in the future, a unittest is
added that cross-checks the default parameters in constants.py and the
hypervisor PARAMETERS dict.

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

Allows to install Red Hat based systems, for example Oracle Linux.
Tested with OEL.

The hypervisor by default offers a device named 'sda'.  If the SCSI
module is already loaded, the disk device can't be created due to naming
conflict, and the disk is not available.  A workaround is to modify the
initrd by removing the scsi driver from it.  This helps, but doesn't
allow to install the OS.

Red Hat's installer, anaconda, runs parted, which tries to execute a
check against /dev/sda and fails.  This makes anaconda think that the
disk is faulty, and not available.  The best way to work around this, is
to declare 'xvda' as the xen disk device.  Red Hat version of parted
package contains a patch which makes parted skip the SCSI test if device
name starts with 'xvd'.

This patch allows to pass -H xen-pvm:blockdev_prefix="xvd" and
successfully run the Red Hat installer.

Signed-off-by: Maciej Bliziński <blizinski@google.com>
[iustin@google.com: added the new parameter to XenHvm PARAMS]
Signed-off-by: Iustin Pop <iustin@google.com>
Reviewed-by: Iustin Pop <iustin@google.com>

Currently, the call_node_info RPC does always check both the VG free
space and the hypervisor information. However, in ⅔ of the uses, we only
care about one or the other. Therefore, we change it so that if any of
the passed parameters is None, we don't perform the respective check. We
also modify its callers to only pass in what they need.

This also helps if the "default" hypervisor is broken and we want to
create an instance for another hypervisor.

With this patch, the duration of this rpc changes from 500ms to 90ms for
a normal LVM+Xen PVM node, when we only require the LVM data; when we
only require the hypervisor data, it doesn't change (as the “xm list”
time is dominant).

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

First change is to replace "range(len(self.NOTIFIERS)))" with
self.NOTIFIERS, as range(len(range(N))) ≡ range(N).

Sencond change is switch from an explicit indexing to use of zip.
Unittests still pass :)

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

This only tries to upload the ssh_known_hosts_file to the online nodes,
eliminating a timeout and warning for offline nodes.

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

In the previous patch where I modified LUSetNodeParams to take ndparams
into account I missed this additional check of parameters in the Logical
Unit. This patch adds ndparams to the recognized parameter lists which
has to be filled out if nothing else is.

Signed-off-by: René Nussbaumer <rn@google.com>
Reviewed-by: Iustin Pop <iustin@google.com>

Signed-off-by: René Nussbaumer <rn@google.com>
Reviewed-by: Iustin Pop <iustin@google.com>

Since we now have just two possibilities (plaint to drbd or drbd to
plain), we can assume (+an assert) that we come from LVM and access
directly logical_id[0] for the VG on which it is stored.

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

Was substituted by _CheckNodesFreeDiskPerVG.

Signed-off-by: Dmitry Chernyak <dmi.chernyak@gmail.com>
Signed-off-by: Iustin Pop <iustin@google.com>
Reviewed-by: Iustin Pop <iustin@google.com>

Also should work disk growing, instance moving and prerequisite checks.

Signed-off-by: Dmitry Chernyak <dmi.chernyak@gmail.com>
[iustin@google.com: small style fixes]
Signed-off-by: Iustin Pop <iustin@google.com>
Reviewed-by: Iustin Pop <iustin@google.com>

Implemented _CheckNodesFreeDiskPerVG and _ComputeDiskSizePerVG functions
for VG-aware iterations.

Signed-off-by: Dmitry Chernyak <dmi.chernyak@gmail.com>
[iustin@google.com: some style fixes (long lines, etc.)]
Signed-off-by: Iustin Pop <iustin@google.com>
Reviewed-by: Iustin Pop <iustin@google.com>

Added "vg" argument to the "gnt-instance ... --disk" syntax specification.
Now is ok to write:

gnt-instance add ... --disk N:size=NNg,vg=VG_NAME ...

But not all internal structures are ready to handle this yet.

Signed-off-by: Dmitry Chernyak <dmi.chernyak@gmail.com>
[iustin@google.com: removed changes to design-2.0.rst]
Signed-off-by: Iustin Pop <iustin@google.com>
Reviewed-by: Iustin Pop <iustin@google.com>

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

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

Currently, the coverage reports include the unittests themselves, and
this skewes unfairly the reports, as the coverage for the tests is very
high (since they all run).

To fix this, we export the ganeti temp dir from run-in-temp-dir, and we
use that to exclude the tests directory. The patch also fixes a but
related to multiple directories to be omitted (--omit a --omit b is
wrong, it needs to be --omit a,b).

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

Signed-off-by: René Nussbaumer <rn@google.com>
Reviewed-by: Michael Hanselmann <hansmi@google.com>

Signed-off-by: René Nussbaumer <rn@google.com>
Reviewed-by: Michael Hanselmann <hansmi@google.com>

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

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

While starting to implement this, I found a number of deficiencies:

- Drop regular expressions. As it turned out, only very few fields for
  instances used them, all of which can easily be turned into static
  field names.
- Use two separate calls with a request and response description each.
- Add forgotten list of fields to query request.
- Add value status for case where a field is not available for an item,
  e.g. NIC 3 for an instance with only one network interface.
- Add "timestamp" field type.
- Updated examples.

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

Extend allocation functions to be aware of different vg names.  Suppose,
vg_name is helded in disk.physical_id[0] (already was) LUCreateInstance
is modified to be aware of new "vg" argument of the disk specification.

cfg.GetVGName() will only be used when no VG is specified.

The "add" and "remove" for plain and drbd instances should be correct.
The "move" for plain instance should be correct.

Signed-off-by: Dmitry Chernyak <dmi.chernyak@gmail.com>
[iustin@google.com: fixed missing parameters in two cases and tabs]
Signed-off-by: Iustin Pop <iustin@google.com>
Reviewed-by: Iustin Pop <iustin@google.com>

Signed-off-by: René Nussbaumer <rn@google.com>
Reviewed-by: Iustin Pop <iustin@google.com>

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

There are two issues: first, it's simpler to assign directly to pt
instead of via another local variable. Second, this code doesn't make
sense:
  if lt < 0:
    break
  pt = max(0, lt)

If lt is indeed < 0, then the code will break. Otherwise, lt ≥ 0, and
max(0, lt) will always be equal to lt. So we remove the extra
assignment.

Signed-off-by: Iustin Pop <iustin@google.com>
Reviewed-by: René Nussbaumer <rn@google.com>

This patch makes a simple timeout test (e.g. testTimeoutKill) go from
13K poll syscalls to just 6. The code didn't take into account that
RunningTimeout talks in seconds while the poller object expects
milliseconds.

Signed-off-by: Iustin Pop <iustin@google.com>
Reviewed-by: René Nussbaumer <rn@google.com>

Currently the test executes "/bin/sh -c 'trap "" TERM …" via the shell,
which means we have two shells involved (the parent one and the child
one, which does the trap). For some weird reason, this works for older
OSes: the first shell exec's the second one, whereas on squeeze it forks
and only then exec. This might be due to changes in Python or bash or
whatever.

In any case, the point is that this is a hacky way to call a shell, so
we change this to simply execute the desired shell directly, without any
intermediate shells (I don't understand why it was written as such
originally).

Signed-off-by: Iustin Pop <iustin@google.com>
Reviewed-by: René Nussbaumer <rn@google.com>

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

Signed-off-by: René Nussbaumer <rn@google.com>
Reviewed-by: Michael Hanselmann <hansmi@google.com>

* devel-2.3:
  Updates NEWS and configure.ac for 2.3.0~rc1
  Update NEWS & configure.ac for the 2.2.2 release
  Fix documentation regarding conversion to drbd
  Fix documentation regarding conversion to drbd
  Reinstall instance: disallow offline secondaries
  QA: check that doubly modifying an OS state is OK
  Fix breakage in OS state modify
  QA: add tests for gnt-cluster modify -B
  LUSetClusterParms: fix validation of beparams
  Add unittests for TemporaryReservationManager
  TempReservationManager: Reserved() doesn't work

Conflicts:
	man/gnt-instance.sgml (ported to the RST version)

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

make check < /dev/null breaks these tests. Until we have change RunCmd
to override stdin, we need to remove them.

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

Further investigations have to be done for merging some of these bits
together with import-export daemon which uses similiar logic.

Signed-off-by: René Nussbaumer <rn@google.com>
Reviewed-by: Michael Hanselmann <hansmi@google.com>

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

* devel-2.2:
  Update NEWS & configure.ac for the 2.2.2 release
  Fix documentation regarding conversion to drbd

Conflicts:
	NEWS         (integrated 2.2 changes)
	configure.ac (kept our version)

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

This imports the 2.1.8 NEWS entry and adds the 2.2.2 one, then updates the
configure.ac version.

Signed-off-by: Iustin Pop <iustin@google.com>
Reviewed-by: René Nussbaumer <rn@google.com>

Signed-off-by: Iustin Pop <iustin@google.com>
Reviewed-by: René Nussbaumer <rn@google.com>

Signed-off-by: Iustin Pop <iustin@google.com>
Reviewed-by: René Nussbaumer <rn@google.com>

Currently, reinstallation of a DRBD instance with the secondary node offline does:

node1# gnt-instance reinstall -f instance1
Waiting for job 139053 for instance1...
Thu Nov 18 01:36:09 2010  - WARNING: Could not prepare block device disk/0 on node node3 (is_primary=False, pass=1): Node is marked offline
Thu Nov 18 01:36:09 2010  - WARNING: Could not shutdown block device disk/0 on node node3: Node is marked offline
Job 139053 for instance1 has failed: Failure: command execution error:
Disk consistency error

Since this fails anyway, let's check the secondary nodes, thus
preventing any modifications to the instance (e.g. OS type change):

node1# gnt-instance reinstall -f instance1
Waiting for job 139058 for instance1...
Job 139058 for instance1 has failed: Failure: prerequisites not met for this operation:
error type: wrong_state, error details:
Instance secondary node offline, cannot reinstall: node3

The patch needs modifications to the _CheckNodeOnline function, in order
to display meaningful messages ("Can't use offline node" would be very
confusing for an instance reinstall, since we didn't select a node
manually).

Signed-off-by: Iustin Pop <iustin@google.com>
Reviewed-by: René Nussbaumer <rn@google.com>

This would have prevented the bug fixed in the previous patch :(

Signed-off-by: Iustin Pop <iustin@google.com>
Reviewed-by: René Nussbaumer <rn@google.com>

I was using the feedback_fn function incorrectly (it doesn't
automatically expand the arguments).

Signed-off-by: Iustin Pop <iustin@google.com>
Reviewed-by: René Nussbaumer <rn@google.com>

* devel-2.2:
  QA: add tests for gnt-cluster modify -B
  LUSetClusterParms: fix validation of beparams

Conflicts:
	lib/cmdlib.py (reverted & applied manually the change)

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