ganeti-local
10 years agoHotplug: cfgupgrade for deprecated fields
Dimitris Aragiorgis [Fri, 12 Jul 2013 12:34:07 +0000 (15:34 +0300)]
Hotplug: cfgupgrade for deprecated fields

Add cfgupgrade so that we remove any deprecated fields hotplug
implementation used in the past.

Signed-off-by: Dimitris Aragiorgis <dimara@grnet.gr>

10 years agoHotplug: Update manpages and NEWS
Dimitris Aragiorgis [Fri, 26 Jul 2013 10:33:38 +0000 (13:33 +0300)]
Hotplug: Update manpages and NEWS

Add --hotplug entry in gnt-instance manpage.
Add NEWS entry for hotplug support.
In both cases mention that hotplug is only supported only the
the KVM Hypervisror and for versions >= 1.0.

Signed-off-by: Dimitris Aragiorgis <dimara@grnet.gr>

10 years agoHotplug: client support
Dimitris Aragiorgis [Fri, 12 Jul 2013 12:33:48 +0000 (15:33 +0300)]
Hotplug: client support

Add --hotplug option. Only used in OpInstanceSetParams.
If this is omitted, modifications become effective after reboot.

Ask user confirmation in case NIC modify + hotplug because it will
be done via removing old NIC (and the corresponding tap) and adding
a new one in the same PCI slot.

Corresponding mods in haskell opcode definitions.

Signed-off-by: Dimitris Aragiorgis <dimara@grnet.gr>

10 years agoHotplug: cmdlib support
Dimitris Aragiorgis [Sat, 25 May 2013 16:13:54 +0000 (19:13 +0300)]
Hotplug: cmdlib support

Hotpluging is done by functions invoked by ApplyContainerMods(). In
order hotpluging to take place the --hotplug option must be passed
otherwise the modifications will take place after reboot.

NIC hotplug supports add, remove and modify. The modify is done by removing
the existing NIC and adding a new one in the same pci slot.

Disk hotplug support add and remove.  Before hotpluging a Disk it
must be assembled. During LUInstanceSetParams() newly created disks
are not added to the instance so ExpandCheckDisks() in
AssembleInstanceDisks() will fail. So we make this check optional
only for this case.

In order to remove a disk (with blockdev_remove) it must be shutdown. So
after unpluging the disk ShutdownDiskInstanceDisks() must be invoked.

For both device types we use the generic RPC call_hotplug_device.

Signed-off-by: Dimitris Aragiorgis <dimara@grnet.gr>

10 years agoHotplug: rpc support
Dimitris Aragiorgis [Thu, 9 Aug 2012 10:59:11 +0000 (13:59 +0300)]
Hotplug: rpc support

Introduce new RPC that eventually invoke hypervisor specific
hotplug functions. In order to be generic it has the following
arguments: device type, action, device, extra info, seq.
Device type can be NIC or DISK, action can be ADD, REMOVE,
device is the NIC or Disk object, extra info is used by Disk
hotplug to point the device path and seq is the device index
(from the master perspective)

If hypervisor does not support hotplug the opcode will fail.

Signed-off-by: Dimitris Aragiorgis <dimara@grnet.gr>

10 years agoIntroduce hotplug methods (hypervisor inteface)
Dimitris Aragiorgis [Fri, 2 Aug 2013 20:00:15 +0000 (23:00 +0300)]
Introduce hotplug methods (hypervisor inteface)

4 new methods: HotAddDevice, HotDelDevice, HotModDevice, HotplugSupported
All these are the hypervisor interface with the backend.
In general they read current runtime info, do the actual hotplug
and update the runtime file.

NIC hot-add:
 - open a tap and get its file descriptor.
 - pass fd with SCM rights (using python-fdsend) via monitor socket
 - create netdev and device with id=kvm_devid and proper pci info

Disk hot-add:
 - create drive with id=kvm_devid
 - create device with id=kvm_devid and corresponding pci info

Use MonitorSocket in _PassTapFd in order to connect to monitor socket
and pass the tap's file descriptor using fdsend.

Signed-off-by: Dimitris Aragiorgis <dimara@grnet.gr>

10 years agoExtract socket related code from QmpMonitor class
Dimitris Aragiorgis [Mon, 5 Aug 2013 10:22:17 +0000 (13:22 +0300)]
Extract socket related code from QmpMonitor class

Separate unix socket related code from QmpMonitor class and
make the latter extend the newly introduced one: MonitorSocket

Signed-off-by: Dimitris Aragiorgis <dimara@grnet.gr>

10 years agoHelper methods that check for hotplug support
Dimitris Aragiorgis [Fri, 2 Aug 2013 19:54:46 +0000 (22:54 +0300)]
Helper methods that check for hotplug support

Hotplug is currently *not* supported for:

 - KVM < 1.0
 - existing devices in the cluster
 - python-fdsend module is not installed (NIC hotplug)
 - chroot (Disk hotplug)
 - security mode other than None (Disk hotplug)

For the above reasons raise HotplugError exception and
let masterd handle it.

TODO: modify CallHotplugCommand() so that it parses
monitor output and reports whether the command
succeeded or not.

Signed-off-by: Dimitris Aragiorgis <dimara@grnet.gr>

10 years agoUse -device with pci info if possible
Dimitris Aragiorgis [Fri, 2 Aug 2013 19:50:00 +0000 (22:50 +0300)]
Use -device with pci info if possible

Extend kvm_cmd with -device option for the case of paravirual
Disks and NICs.

Signed-off-by: Dimitris Aragiorgis <dimara@grnet.gr>

10 years agoHepler methods for PCI slots and device ids
Dimitris Aragiorgis [Mon, 5 Aug 2013 15:26:48 +0000 (18:26 +0300)]
Hepler methods for PCI slots and device ids

Device naming:
QEMU monitor expects devices to be uniquely named. Device ids derive
from the following function:
kvm_devid = <device_type>-<part of uuid>-pci-<pci_slot>
Device ids must be reproduce-able when we want to remove them.
For that reason we store the pci slot inside the runtime file and
in case we want to remove a device we obtain its pci slot by
parsing the corresponding runtime enrty and matching the device
by its uuid.

Finding the PCI slot:
For newly added devices Hypervisor parses existing PCI allocations
(via _GetFreePCISlot() and eventually ``info pci`` monitor
command) and decides the PCI slot to plug in the device. During
instance startup hypervisor invokes _UpdatePCISlots() for every
device of the instance.  Initial PCI reservations derive from KVM
default setup, that allocates 4 slots for devices other than disks
and NICs.

Signed-off-by: Dimitris Aragiorgis <dimara@grnet.gr>

10 years agoPrepare block_devices runtime entry
Dimitris Aragiorgis [Mon, 5 Aug 2013 15:23:23 +0000 (18:23 +0300)]
Prepare block_devices runtime entry

With this patch we add another entry in runtime files along with
kvm_cmd, kvm_nics, and hvparams. block_devices that used to be
encapsulated inside kvm_cmd, live now separately just like nics do
but in tupples of (L{objects.Disk}, dev_path).

Introduce also _GetExistingDeviceInfo() helper function to search
for runtime entries. This is going to be useful later in hotplug
methods.

Define new exception: HotplugError.

Signed-off-by: Dimitris Aragiorgis <dimara@grnet.gr>

10 years agoPrepare kvm options to support -device for disks
Dimitris Aragiorgis [Fri, 2 Aug 2013 16:44:29 +0000 (19:44 +0300)]
Prepare kvm options to support -device for disks

Newer kvm version support -device driver,... option where available
drivers can be found by running kvm -device ? command. Currently
ganeti uses `if` property of driver to define the disk type to use
which is also passed as hvparam. This patch prepares the use of
-device option only for paravirtual disks. To support other drivers
as well we must create a mapping between the values passed now as
disk_type hvparam and existing drivers. Note that if -device is used
`if` property of `drive` option should be `none`.

Signed-off-by: Dimitris Aragiorgis <dimara@grnet.gr>

10 years agoIntroduce _GenerateKVMBlockDevicesOptions()
Dimitris Aragiorgis [Fri, 2 Aug 2013 16:37:02 +0000 (19:37 +0300)]
Introduce _GenerateKVMBlockDevicesOptions()

Simply move logic that generates block device related
options out of _GenerateKVMRuntime().

Signed-off-by: Dimitris Aragiorgis <dimara@grnet.gr>

10 years agoAdd hotplug related constants
Dimitris Aragiorgis [Sat, 21 Sep 2013 08:32:53 +0000 (11:32 +0300)]
Add hotplug related constants

Signed-off-by: Dimitris Aragiorgis <dimara@grnet.gr>

10 years agoHotplug: objects and constants additions
Dimitris Aragiorgis [Thu, 9 Aug 2012 10:54:18 +0000 (13:54 +0300)]
Hotplug: objects and constants additions

Add pci slot in NIC/Disk objects. This slot will be used only
by hypervisor code. Currently only KVM will use it and store it
temporarily in runtime files.
Add HOTPLUG_* constants to define device types an hotplug actions.

Signed-off-by: Dimitris Aragiorgis <dimara@grnet.gr>

10 years agoVersion bump for 2.8.2 v2.8.2
Michele Tartara [Wed, 6 Nov 2013 12:26:24 +0000 (12:26 +0000)]
Version bump for 2.8.2

Signed-off-by: Michele Tartara <mtartara@google.com>
Reviewed-by: Thomas Thrainer <thomasth@google.com>

10 years agoUpdate NEWS file for 2.8.2 release
Michele Tartara [Wed, 6 Nov 2013 12:25:16 +0000 (12:25 +0000)]
Update NEWS file for 2.8.2 release

Add a section in the file for the new upcoming release.

Signed-off-by: Michele Tartara <mtartara@google.com>
Reviewed-by: Thomas Thrainer <thomasth@google.com>

10 years agoDRBD: ensure peers are UpToDate for dual-primary
Apollon Oikonomopoulos [Tue, 5 Nov 2013 14:30:45 +0000 (16:30 +0200)]
DRBD: ensure peers are UpToDate for dual-primary

DrbdAttachNet supports both, normal primary/secondary node operation, and
(during live migration) dual-primary operation. When resources are newly
attached, we poll until we find all of them in connected or syncing operation.

Although aggressive, this is enough for primary/secondary operation, because
the primary/secondary role is not changed from within DrbdAttachNet. However,
in the dual-primary ("multimaster") case, both peers are subsequently upgraded
to the primary role.  If - for unspecified reasons - both disks are not
UpToDate, then a resync may be triggered after both peers have switched to
primary, causing the resource to disconnect:

  kernel: [1465514.164009] block drbd2: I shall become SyncTarget, but I am
    primary!
  kernel: [1465514.171562] block drbd2: ASSERT( os.conn == C_WF_REPORT_PARAMS )
    in /build/linux-rrsxby/linux-3.2.51/drivers/block/drbd/drbd_receiver.c:3245

This seems to be extremely racey and is possibly triggered by some underlying
network issues (e.g. high latency), but it has been observed in the wild. By
logging the DRBD resource state in the old secondary, we managed to see a
resource getting promoted to primary while it was:

  WFSyncUUID Secondary/Primary Outdated/UpToDate

We fix this by explicitly waiting for "Connected" cstate and
"UpToDate/UpToDate" disks, as advised in [1]:

  "For this purpose and scenario,
   you only want to promote once you are Connected UpToDate/UpToDate."

[1] http://lists.linbit.com/pipermail/drbd-user/2013-July/020173.html

Signed-off-by: Apollon Oikonomopoulos <apoikos@gmail.com>
Signed-off-by: Michele Tartara <mtartara@google.com>
Reviewed-by: Michele Tartara <mtartara@google.com>
Reviewed-by: Klaus Aehlig <aehlig@google.com>

10 years agoImprove error message for replace-disks
Michele Tartara [Mon, 4 Nov 2013 15:20:07 +0000 (15:20 +0000)]
Improve error message for replace-disks

In some conditions, replace-disks will fail if the disks are not properly
activated. Improve the error message suggesting to run activate-disks before
executing replace-disks.

Fixes Issue 606.

Signed-off-by: Michele Tartara <mtartara@google.com>
Reviewed-by: Klaus Aehlig <aehlig@google.com>

10 years agoAdd all dependencies for confd as test dependencies
Klaus Aehlig [Wed, 30 Oct 2013 10:13:48 +0000 (11:13 +0100)]
Add all dependencies for confd as test dependencies

Since our tests pull in confd as a dependency, all build dependencies
for confd are also necessary to run the tests.

Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Michele Tartara <mtartara@google.com>

10 years agoAdd snap-server to the test-relevenat packages
Klaus Aehlig [Tue, 29 Oct 2013 15:09:14 +0000 (16:09 +0100)]
Add snap-server to the test-relevenat packages

While snap-server is only needed for the optional monitoring daemon,
some tests, notably those testing these optional features, still depend
on it. So, if snap-server is missing, the Haskell tests should not be
run, as they cannot even be build.

Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Michele Tartara <mtartara@google.com>

10 years agoPlacate warnings on ganeti.outils_unittest.py
Santi Raffa [Wed, 23 Oct 2013 14:20:24 +0000 (16:20 +0200)]
Placate warnings on ganeti.outils_unittest.py

Change ovf.py from using 'if foo:' to 'if foo is not None:' checks as
suggested by the FutureWarnings the OVF library raises otherwise.
This fixes issue 557.

Signed-off-by: Santi Raffa <rsanti@google.com>
Reviewed-by: Jose Lopes <jabolopes@google.com>

10 years agoVersion bump for 2.8.1 v2.8.1
Michele Tartara [Thu, 17 Oct 2013 10:36:44 +0000 (12:36 +0200)]
Version bump for 2.8.1

Also, update the NEWS file accordingly.

Signed-off-by: Michele Tartara <mtartara@google.com>
Reviewed-by: Klaus Aehlig <aehlig@google.com>

10 years agoVerify that ConfD is running after master-failover
Helga Velroyen [Wed, 16 Oct 2013 11:11:18 +0000 (13:11 +0200)]
Verify that ConfD is running after master-failover

This patch adds additional 'gnt-node list' commands to
the testing of 'gnt-cluster master-failover' in order to
test if ConfD (or LuxiD) is still running after a
master-failover.

Signed-off-by: Helga Velroyen <helgav@google.com>
Reviewed-by: Klaus Aehlig <aehlig@google.com>

10 years agodaemon-util: handle luxid in {start,stop}_master()
Apollon Oikonomopoulos [Wed, 16 Oct 2013 09:17:23 +0000 (12:17 +0300)]
daemon-util: handle luxid in {start,stop}_master()

Luxid was not handled in start_master() and stop_master() at all. As a result,
during a master-failover, luxid would be left running on the old master and
would not start on the new master, leaving the cluster without management until
luxid was manually started.

Signed-off-by: Apollon Oikonomopoulos <apoikos@gmail.com>
Signed-off-by: Michele Tartara <mtartara@google.com>
Reviewed-by: Michele Tartara <mtartara@google.com>

10 years agoFix typo in storage.FileStorage docstring
Santi Raffa [Tue, 15 Oct 2013 08:13:28 +0000 (10:13 +0200)]
Fix typo in storage.FileStorage docstring

Signed-off-by: Santi Raffa <rsanti@google.com>
Reviewed-by: Jose Lopes <jabolopes@google.com>

10 years agoFix path for serial file
Klaus Aehlig [Wed, 16 Oct 2013 07:32:02 +0000 (09:32 +0200)]
Fix path for serial file

It is actually located inside the queue directory.

Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Helga Velroyen <helgav@google.com>

10 years agoImprove harep documentation
Michele Tartara [Thu, 10 Oct 2013 14:03:44 +0000 (14:03 +0000)]
Improve harep documentation

Add a more complete description of how the tool works.

Signed-off-by: Michele Tartara <mtartara@google.com>
Reviewed-by: Klaus Aehlig <aehlig@google.com>

10 years agoFix typo in walkthrough documentation
Jose A. Lopes [Tue, 8 Oct 2013 11:26:47 +0000 (13:26 +0200)]
Fix typo in walkthrough documentation

Fix typo in walkthrough documentation, where 'instance3' should read
'instance1' in order to be consistent with the output below in the
example.

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

10 years agoDon't attemp ipv6 ssh in case of ipv4 cluster
Costas Drogos [Thu, 3 Oct 2013 09:37:14 +0000 (12:37 +0300)]
Don't attemp ipv6 ssh in case of ipv4 cluster

In case of a cluster with primary-ip-version=4, there is no need
for cluster to try ipv6 ssh connections which may timeout.
So append '-4' on ssh if cluster is ipv4-only

Signed-off-by: Costas Drogos <costas.drogos@gmail.com>
Signed-off-by: Michele Tartara <mtartara@google.com>
Reviewed-by: Michele Tartara <mtartara@google.com>

10 years agoFix outdated documentation for users-setup
Michele Tartara [Mon, 7 Oct 2013 07:51:30 +0000 (07:51 +0000)]
Fix outdated documentation for users-setup

The --yes-do-it parameter was not properly documented yet. Add it to the
administration guide.

Signed-off-by: Michele Tartara <mtartara@google.com>
Reviewed-by: Klaus Aehlig <aehlig@google.com>

10 years agoFix wrong release date in the NEWS file v2.8.0
Michele Tartara [Fri, 27 Sep 2013 13:02:14 +0000 (15:02 +0200)]
Fix wrong release date in the NEWS file

Signed-off-by: Michele Tartara <mtartara@google.com>
Reviewed-by: Klaus Aehlig <aehlig@google.com>

10 years agoVersion bump for 2.8.0
Michele Tartara [Fri, 27 Sep 2013 12:32:58 +0000 (14:32 +0200)]
Version bump for 2.8.0

Also, update the NEWS file accordingly.

Signed-off-by: Michele Tartara <mtartara@google.com>
Reviewed-by: Klaus Aehlig <aehlig@google.com>

10 years agoAdd daemon split design doc
Michele Tartara [Tue, 6 Aug 2013 15:19:49 +0000 (15:19 +0000)]
Add daemon split design doc

This describes the future planned structure of Ganeti daemons.

Signed-off-by: Michele Tartara <mtartara@google.com>
Reviewed-by: Guido Trotter <ultrotter@google.com>

10 years agoMerge branch 'stable-2.7' into stable-2.8
Michele Tartara [Thu, 26 Sep 2013 11:32:57 +0000 (13:32 +0200)]
Merge branch 'stable-2.7' into stable-2.8

Conflicts are trivially solved.

* stable-2.7
  Version bump for 2.7.2 (updated NEWS file)

Signed-off-by: Michele Tartara <mtartara@google.com>
Reviewed-by: Guido Trotter <ultrotter@google.com>

Conflicts:
NEWS
configure.ac

10 years agoVersion bump for 2.7.2 v2.7.2
Michele Tartara [Thu, 26 Sep 2013 07:53:10 +0000 (09:53 +0200)]
Version bump for 2.7.2

Change the version number and the NEWS file for the 2.7.2 bugfix release.

Signed-off-by: Michele Tartara <mtartara@google.com>
Reviewed-by: Thomas Thrainer <thomasth@google.com>

10 years agoAdd additional tests for utils.Retry
Klaus Aehlig [Fri, 20 Sep 2013 11:45:47 +0000 (13:45 +0200)]
Add additional tests for utils.Retry

Also have some tests where the time the various components
(inspecting the time function, calling the function, etc) have
a non-trivial time. Also, have two tests demonstrating that
the number of retries actually depends on the amount of time
the function takes to complete.

Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Thomas Thrainer <thomasth@google.com>

10 years agoMake retry tests independent of actual time
Klaus Aehlig [Thu, 19 Sep 2013 15:05:29 +0000 (17:05 +0200)]
Make retry tests independent of actual time

Even in the tests, real time is used. While, generally, the assumptions
about execution time are pretty safe, in some rare circumstances, e.g.,
on machines with extremely heavy load they do not hold true, thus rendering
the tests flaky. Fix this, by mocking time.

Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Thomas Thrainer <thomasth@google.com>

10 years agoFix corner-case in handling of remaining retry time
Klaus Aehlig [Thu, 19 Sep 2013 16:12:21 +0000 (18:12 +0200)]
Fix corner-case in handling of remaining retry time

Consider a remaining time of 0 as already timed out. Otherwise,
there is no guarantee that calling utils.Retry with timeout 0
will call the function precisely once; it might run in time
shorter than the resolution of timer.

Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Thomas Thrainer <thomasth@google.com>

10 years agoPerform proper cleanup on termination of Haskell daemons
Michele Tartara [Wed, 18 Sep 2013 13:38:18 +0000 (15:38 +0200)]
Perform proper cleanup on termination of Haskell daemons

Haskell deamons did not perform proper cleanup at termination. There was no code
for removing the pid file, and the code in LuxiD for removing the unix socket
file was not working, because it is implemented with a "finally" statement,
which is executed only when the main loop of the daemon is exited (either
normally, or through an exception), but not when it is terminated by a SIGTERM.

This commit adds a proper handler for SIGTERM, which transforms it into a
successful termination exception. This allows both the newly added cleanup code
for pid files and the unmodified cleanup code for unix sockets to be executed.

Fixes Issue 581 and 582.

Signed-off-by: Michele Tartara <mtartara@google.com>
Reviewed-by: Klaus Aehlig <aehlig@google.com>

10 years agoVersion bump for 2.8 rc3 v2.8.0rc3
Michele Tartara [Tue, 17 Sep 2013 08:43:02 +0000 (10:43 +0200)]
Version bump for 2.8 rc3

Also, add to the NEWS file a couple of items that were still missing.

Signed-off-by: Michele Tartara <mtartara@google.com>
Reviewed-by: Thomas Thrainer <thomasth@google.com>

10 years agoMerge branch 'stable-2.7' into stable-2.8
Klaus Aehlig [Mon, 16 Sep 2013 11:52:45 +0000 (13:52 +0200)]
Merge branch 'stable-2.7' into stable-2.8

* stable-2.7
  Fix incorrect manpage reference to htools

Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Michele Tartara <mtartara@google.com>

10 years agoFix incorrect manpage reference to htools
Martin Zobel-Helas [Mon, 16 Sep 2013 11:42:46 +0000 (13:42 +0200)]
Fix incorrect manpage reference to htools

the manpages for hcheck, hinfo and hroller have an incorrect reference
to htools(7) which does not exist. This patch links them to htools(1).

This bug also fixes Debian BTS#721824 (http://bugs.debian.org/721824)

Signed-off-by: Martin Zobel-Helas <zobel@debian.org>
Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Klaus Aehlig <aehlig@google.com>

10 years agoAdd a default to file-driver when unspecified over RAPI
Michele Tartara [Mon, 16 Sep 2013 11:06:31 +0000 (11:06 +0000)]
Add a default to file-driver when unspecified over RAPI

The file-driver value, used by file-based instances, had a default value when
an instance was being created over the CLI, but not when the instance was
created through the RAPI.

This patch introduces a default value for the remote API and, while doing so,
also centralizes in a single place the definition of such a value, for both the
interfaces, in order to remove the possibility for them to be out of sync.

Fixes Issue 571.

Signed-off-by: Michele Tartara <mtartara@google.com>
Reviewed-by: Klaus Aehlig <aehlig@google.com>

10 years agoRemove typo from RAPI documentation
Michele Tartara [Fri, 13 Sep 2013 09:27:51 +0000 (09:27 +0000)]
Remove typo from RAPI documentation

"be used instead." was repeated twice.

Signed-off-by: Michele Tartara <mtartara@google.com>
Reviewed-by: Klaus Aehlig <aehlig@google.com>

10 years agoDescribe the files in doc/users in NEWS
Klaus Aehlig [Thu, 12 Sep 2013 09:38:24 +0000 (11:38 +0200)]
Describe the files in doc/users in NEWS

While only a small feature, the files in doc/users can influence
the way ganeti is packaged. Hence it should be mentioned in the
news file.

Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Michele Tartara <mtartara@google.com>

10 years agoUpdate the upgrade procedure
Klaus Aehlig [Thu, 12 Sep 2013 09:08:34 +0000 (11:08 +0200)]
Update the upgrade procedure

Now that users-setup asks for confirmation, explain in the
procedure how asking for this confirmation can be skipped.
This skipping is particularly important for users following
the recommendation of using the distributed shell (dsh).

Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Michele Tartara <mtartara@google.com>

10 years agoMake tools/users-setup explain its actions first
Klaus Aehlig [Wed, 11 Sep 2013 09:00:15 +0000 (11:00 +0200)]
Make tools/users-setup explain its actions first

Before doing any changes, make tools/users-setup explain its
actions first, and ask the user for confirmation. If the first
argument is --yes-do-it, this step will be skipped.

Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Michele Tartara <mtartara@google.com>

10 years agoGenerate tools/users-setup from descriptions
Klaus Aehlig [Wed, 11 Sep 2013 08:27:23 +0000 (10:27 +0200)]
Generate tools/users-setup from descriptions

Now that we have a machine readable description of the users
to be added to the system, use it to generate the tools/users-setup
script, so that we again have only one authoritative source of
the user and group requirements. This also has the advantage of
of avoiding duplicate additions and attemps add a user root.

Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Michele Tartara <mtartara@google.com>

10 years agoProvide ganeti user and group data in easy-to-parse form
Klaus Aehlig [Tue, 10 Sep 2013 16:00:23 +0000 (18:00 +0200)]
Provide ganeti user and group data in easy-to-parse form

Upon build time, also generate, in the directory doc/users, files
containing the groups, the users with their primary group, and the
additional groupmemberships to be added. In this way, packages can use
their own way of adding users to the system, instead of telling the
adiministrator to run tools/users-setup.

Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Michele Tartara <mtartara@google.com>

10 years agoRemove dsahostkeypub during config downgrade
Apollon Oikonomopoulos [Sun, 1 Sep 2013 20:37:08 +0000 (23:37 +0300)]
Remove dsahostkeypub during config downgrade

Remove the newly-added DSA host public key field from the configuration file
when downgrading to 2.7, if present. Otherwise masterd will refuse to start.

Signed-off-by: Apollon Oikonomopoulos <apoikos@gmail.com>
Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Klaus Aehlig <aehlig@google.com>

10 years agoMark the DSA host pubkey as optional
Apollon Oikonomopoulos [Sun, 1 Sep 2013 20:36:29 +0000 (23:36 +0300)]
Mark the DSA host pubkey as optional

Commit a9542a4 introduced support for DSA SSH keys. However, the dsahostkeypub
field added to the config is not marked as optional in the Haskell components.
As a result, luxid thinks the config file is corrupt and refuses to start. We
fix this by marking the dsahostkeypub as an optionalField.

This fixes issue 560.

Signed-off-by: Apollon Oikonomopoulos <apoikos@gmail.com>
Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Klaus Aehlig <aehlig@google.com>

10 years agoFix documentation for gnt-node evacuate -p option
Leon Handreke [Wed, 28 Aug 2013 15:10:49 +0000 (16:10 +0100)]
Fix documentation for gnt-node evacuate -p option

Update the documentation to remove the claim that the --primary-only
option results in data copies while in fact it only fails over/migrates
the all primary instances.

Signed-off-by: Leon Handreke <lhandreke@google.com>
Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Klaus Aehlig <aehlig@google.com>

10 years agoNEWS update and version bump for 2.8 rc2 v2.8.0rc2
Michele Tartara [Tue, 27 Aug 2013 07:59:14 +0000 (09:59 +0200)]
NEWS update and version bump for 2.8 rc2

Latest modifications and fixes added to the NEWS file.

Signed-off-by: Michele Tartara <mtartara@google.com>
Reviewed-by: Guido Trotter <ultrotter@google.com>

10 years agoAdd unit test for GetLinuxNodeInfo
Michele Tartara [Thu, 22 Aug 2013 08:53:07 +0000 (08:53 +0000)]
Add unit test for GetLinuxNodeInfo

The function was not unit tested previously.

Signed-off-by: Michele Tartara <mtartara@google.com>
Reviewed-by: Helga Velroyen <helgav@google.com>

10 years agoPrepare GetLinuxNodeInfo for testing
Michele Tartara [Thu, 22 Aug 2013 08:31:45 +0000 (10:31 +0200)]
Prepare GetLinuxNodeInfo for testing

The names of the input files are not hardcoded anymore, but they are now default
parameters.

No changes for the callers, but it's now possible to test the function more
easily.

Signed-off-by: Michele Tartara <mtartara@google.com>
Reviewed-by: Helga Velroyen <helgav@google.com>

10 years agoMake the DRBD collector more failure-resilient
Michele Tartara [Wed, 21 Aug 2013 14:17:38 +0000 (14:17 +0000)]
Make the DRBD collector more failure-resilient

If information about instances is not available, just log the error and
continue without it.

Signed-off-by: Michele Tartara <mtartara@google.com>
Reviewed-by: Klaus Aehlig <aehlig@google.com>

10 years agoAdd function to unwrap Results logging failures
Michele Tartara [Wed, 21 Aug 2013 14:15:30 +0000 (14:15 +0000)]
Add function to unwrap Results logging failures

Add logWarningIfBad, a utility function similar to exitIfBad, that logs a
warning and returns a default value instead of just crashing the program if
the unpacked value is Bad.

Signed-off-by: Michele Tartara <mtartara@google.com>
Reviewed-by: Klaus Aehlig <aehlig@google.com>

10 years agoAllow instance mods if only osparams are given
Thomas Thrainer [Tue, 20 Aug 2013 10:37:08 +0000 (12:37 +0200)]
Allow instance mods if only osparams are given

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

(cherry picked from commit 5eae613c2e1e65101c5d5f7d2e8ffd7cc6edc7d5)

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

10 years agoFix harep manpage title
Apollon Oikonomopoulos [Fri, 16 Aug 2013 11:58:03 +0000 (14:58 +0300)]
Fix harep manpage title

The harep manpage title underline was one character short, leading in
pandoc producing an invalid manpage header.

Signed-off-by: Apollon Oikonomopoulos <apoikos@gmail.com>
Reviewed-by: Thomas Thrainer <thomasth@google.com>

10 years agoUse FQDN to check master node status
Thomas Thrainer [Tue, 13 Aug 2013 07:50:42 +0000 (09:50 +0200)]
Use FQDN to check master node status

The master node name in SS conf is stored as FQDN, so also use the FQDN
on each node to check if it is the master node.

This fixes issue 551.

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

10 years agoMerge branch 'stable-2.7' into stable-2.8
Thomas Thrainer [Wed, 7 Aug 2013 09:43:44 +0000 (11:43 +0200)]
Merge branch 'stable-2.7' into stable-2.8

* stable-2.7
  Fix documentation of gnt-instance batch-create
  Support multi instance allocs without iallocator
  Document --no-conflicts-check for gnt-network
  Check disk template in right dict when copying
  ganeti.backend_unittest: chmod restricted cmd dir
  Fix formatting of tuple in error message
  Fix 'prcre' typo in INSTALL
  Format gnt-network info group output
  Add 2.7.2 NEWS entry

Conflicts:
INSTALL
      (trivial)
NEWS
      (added entry to unreleased 2.8 section with remark that it was
       merged from 2.7)
lib/cmdlib.py
      (followed cmdlib split)

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

10 years agoFix batch creation of instances
Thomas Thrainer [Mon, 5 Aug 2013 09:06:13 +0000 (11:06 +0200)]
Fix batch creation of instances

During batch creation of instances the 'reason' field in opcodes is set
to None (but the field is present). This caused problems when adding a
reasons to the reason trail.

Setting the default value for the 'reason' field to the empty list fixes
this issue.

Signed-off-by: Thomas Thrainer <thomasth@google.com>
Reviewed-by: Michele Tartara <mtartara@google.com>

10 years agoFix documentation of gnt-instance batch-create
Thomas Thrainer [Mon, 5 Aug 2013 10:50:37 +0000 (12:50 +0200)]
Fix documentation of gnt-instance batch-create

The documentation of gnt-instance batch-create was outdated and
contained invalid examples. This patch corrects the man page and
includes working example inputs.

Signed-off-by: Thomas Thrainer <thomasth@google.com>
Reviewed-by: Michele Tartara <mtartara@google.com>

10 years agoSupport multi instance allocs without iallocator
Thomas Thrainer [Mon, 5 Aug 2013 10:11:21 +0000 (12:11 +0200)]
Support multi instance allocs without iallocator

If all instances in the multi allocation request have already their
primary and secondary node set, there is no need for an iallocator. Thus
don't require it in this case and omit the call to it all together.

Signed-off-by: Thomas Thrainer <thomasth@google.com>
Reviewed-by: Michele Tartara <mtartara@google.com>

10 years agoSupport DSA SSH keys in bootstrap
Thomas Thrainer [Tue, 6 Aug 2013 15:10:15 +0000 (17:10 +0200)]
Support DSA SSH keys in bootstrap

As outlined in issue 338, Ganeti failed to initialize a cluster if no
RSA SSH key is present on the master node. This patch extends Ganetis
support to DSA keys, so clusters with only DSA keys are possible now.

This fixes issue 338.

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

10 years agoInclude VCS version in `gnt-cluster version`
Thomas Thrainer [Tue, 6 Aug 2013 10:54:15 +0000 (12:54 +0200)]
Include VCS version in `gnt-cluster version`

Also print the VCS version in the output of `gnt-cluster version`. This
makes the VCS version also available over RAPI, etc.

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

10 years agoDocument 'viridian' hypervisor parameter
Thomas Thrainer [Tue, 6 Aug 2013 08:09:10 +0000 (10:09 +0200)]
Document 'viridian' hypervisor parameter

List the 'viridian' hypervisor parameter (valid for Xen HVM) in the
gnt-instance manpage.

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

10 years agoSupport 'viridian' parameter in Xen HVM
Thomas Thrainer [Tue, 6 Aug 2013 08:04:58 +0000 (10:04 +0200)]
Support 'viridian' parameter in Xen HVM

This parameter is required to prevent bluescreens in Windows instances.

This fixes issue 233.

Signed-off-by: Heiko Baumann <heibau@googlemail.com>
Signed-off-by: Thomas Thrainer <thomasth@google.com>
Reviewed-by: Helga Velroyen <helgav@google.com>

10 years agoDocument --no-conflicts-check for gnt-network
Thomas Thrainer [Mon, 5 Aug 2013 07:46:30 +0000 (09:46 +0200)]
Document --no-conflicts-check for gnt-network

gnt-network add and connect do support the --no-conflicts-check option.
This is now also documented in the man pages.

This fixes issue 526.

Signed-off-by: Thomas Thrainer <thomasth@google.com>
Reviewed-by: Michele Tartara <mtartara@google.com>

10 years agoVersion bump for 2.8.0 rc1 v2.8.0rc1
Michele Tartara [Wed, 31 Jul 2013 13:21:28 +0000 (13:21 +0000)]
Version bump for 2.8.0 rc1

Update NEWS file and version number.

Signed-off-by: Michele Tartara <mtartara@google.com>
Reviewed-by: Guido Trotter <ultrotter@google.com>

10 years agoAdd Make to the list of dependencies
Michele Tartara [Fri, 2 Aug 2013 07:40:44 +0000 (07:40 +0000)]
Add Make to the list of dependencies

Make is a required dependency, but it was not specified in the install guide.
This commit adds it.

Signed-off-by: Michele Tartara <mtartara@google.com>
Reviewed-by: Thomas Thrainer <thomasth@google.com>

10 years agoCheck for luxid permission during verify config
Thomas Thrainer [Thu, 1 Aug 2013 14:43:40 +0000 (16:43 +0200)]
Check for luxid permission during verify config

confd no longer needs access to the noded certificate file, but luxid
does. Change the check to use the right user.

Signed-off-by: Thomas Thrainer <thomasth@google.com>
Reviewed-by: Michele Tartara <mtartara@google.com>

10 years agoAdd monitoring-query-format.rst to $(docinput)
Apollon Oikonomopoulos [Thu, 1 Aug 2013 09:45:20 +0000 (12:45 +0300)]
Add monitoring-query-format.rst to $(docinput)

Commit 431ff2c10 split the monitoring system documentation in two parts but did
not include the new monitoring-query-format.rst to $(docinput). As a result,
current tarballs of 2.8 do not ship doc/monitoring-query-format.rst.

Signed-off-by: Apollon Oikonomopoulos <apoikos@gmail.com>
Signed-off-by: Michele Tartara <mtartara@google.com>
Reviewed-by: Michele Tartara <mtartara@google.com>

10 years agoFix documentation building rules
Michele Tartara [Thu, 1 Aug 2013 08:41:04 +0000 (10:41 +0200)]
Fix documentation building rules

Implement the correct behavior to be kept when sphinx-build is not installed.
It was already correctly checked, but instead of just disabling the
documentation generation, an error message was printed, having "make" fail.

Signed-off-by: Michele Tartara <mtartara@google.com>
Reviewed-by: Thomas Thrainer <thomasth@google.com>

10 years agoCheck disk template in right dict when copying
Thomas Thrainer [Wed, 31 Jul 2013 13:41:43 +0000 (15:41 +0200)]
Check disk template in right dict when copying

Due to the structure of the code this condition can't possibly be true.
We have to look in the new_diskparams dict instead, otherwise it'd be
possible to try to update a non-existing entry.

(The same patch is in stable-2.7 as 106441d already).

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

10 years agoCheck disk template in right dict when copying
Thomas Thrainer [Wed, 31 Jul 2013 14:16:39 +0000 (16:16 +0200)]
Check disk template in right dict when copying

Due to the structure of the code this condition can't possibly be true.
We have to look in the new_diskparams dict instead, otherwise it'd be
possible to try to update a non-existing entry.

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

10 years agoganeti.backend_unittest: chmod restricted cmd dir
Apollon Oikonomopoulos [Wed, 31 Jul 2013 14:30:29 +0000 (17:30 +0300)]
ganeti.backend_unittest: chmod restricted cmd dir

TestVerifyRestrictedCmdDirectory.testNormal implicitly relies on the current
umask to check the behaviour of backend._VerifyRestrictedCmdDirectory. However,
when run under a more relaxed umask (as in the case of the Debian buildd's
using sbuild and - apparently - umask 0002), it fails.

Since this kind of failure is already tested for previously and since we should
check that a "known-normal" directory passes the test, we explicitly chmod()
the temporary directory to 0755.

Signed-off-by: Apollon Oikonomopoulos <apoikos@gmail.com>
Reviewed-by: Guido Trotter <ultrotter@google.com>

10 years agoFix formatting of tuple in error message
Thomas Thrainer [Wed, 31 Jul 2013 12:58:29 +0000 (14:58 +0200)]
Fix formatting of tuple in error message

The missing str() call caused to raise another exception than the wanted
one. Transforming the tuple to string fixes this problem.

Signed-off-by: Thomas Thrainer <thomasth@google.com>
Reviewed-by: Michele Tartara <mtartara@google.com>

10 years agoMark the UUID design doc as implemented
Michele Tartara [Wed, 31 Jul 2013 09:33:55 +0000 (09:33 +0000)]
Mark the UUID design doc as implemented

The code implementing the UUID design document is included in Ganeti 2.8.

Signed-off-by: Michele Tartara <mtartara@google.com>
Reviewed-by: Guido Trotter <ultrotter@google.com>

10 years agoClarify that cmdlib/ must be removed downgrading
Michele Tartara [Wed, 31 Jul 2013 08:54:14 +0000 (08:54 +0000)]
Clarify that cmdlib/ must be removed downgrading

Downgrading from 2.8 to a previous version will fail if the new cmdlib/
directory is not removed.

Signed-off-by: Michele Tartara <mtartara@google.com>
Reviewed-by: Guido Trotter <ultrotter@google.com>

10 years agoFix formatting error in the UPGRADE document
Michele Tartara [Wed, 31 Jul 2013 08:53:27 +0000 (08:53 +0000)]
Fix formatting error in the UPGRADE document

An item of the enumerated list was not rendered correctly.

Signed-off-by: Michele Tartara <mtartara@google.com>
Reviewed-by: Guido Trotter <ultrotter@google.com>

10 years agoMake build_chroot compatible with Squeeze
Michele Tartara [Tue, 30 Jul 2013 15:17:11 +0000 (15:17 +0000)]
Make build_chroot compatible with Squeeze

Modify build_chroot to make it compatible with debian Squeeze. This
is done by using a function instead of a alias for the command that
was failing.

Fixes Issue 507.

Signed-off-by: Michele Tartara <mtartara@google.com>
Reviewed-by: Thomas Thrainer <thomasth@google.com>

10 years agoUpdate security doc and NEWS for luxid
Thomas Thrainer [Tue, 30 Jul 2013 07:18:14 +0000 (09:18 +0200)]
Update security doc and NEWS for luxid

Document the split of luxid from confd in security.rst and in the NEWS
file.

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

10 years agoAdd --cleanup to instance failover manpage
Michele Tartara [Tue, 30 Jul 2013 13:02:14 +0000 (13:02 +0000)]
Add --cleanup to instance failover manpage

Describe the new parameter in the manpage of gnt-instance.

Signed-off-by: Michele Tartara <mtartara@google.com>
Reviewed-by: Guido Trotter <ultrotter@google.com>

10 years agoUpdate NEWS with failover cleanup
Michele Tartara [Tue, 30 Jul 2013 12:06:22 +0000 (12:06 +0000)]
Update NEWS with failover cleanup

Describe the new feature in the NEWS file.

Signed-off-by: Michele Tartara <mtartara@google.com>
Reviewed-by: Guido Trotter <ultrotter@google.com>

10 years agoAdd cleanup parameter to instance failover
Michele Tartara [Tue, 30 Jul 2013 11:59:52 +0000 (11:59 +0000)]
Add cleanup parameter to instance failover

Most of the code is shared with instance migrate, so we actually only need
to add the parameter and pass its value along the the common code.

Also, tests and harep are updated to support the right set of options to
the "failover" opcode.

Signed-off-by: Michele Tartara <mtartara@google.com>
Reviewed-by: Guido Trotter <ultrotter@google.com>

10 years agoFix 'prcre' typo in INSTALL
Guido Trotter [Tue, 30 Jul 2013 09:48:35 +0000 (11:48 +0200)]
Fix 'prcre' typo in INSTALL

Thanks to "Izhar ul Hassan" for reporting the issue.

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

10 years agoUpdate NEWS file for 2.8 wrt vnet_hdr KVM hvparam
Stratos Psomadakis [Mon, 29 Jul 2013 14:14:49 +0000 (17:14 +0300)]
Update NEWS file for 2.8 wrt vnet_hdr KVM hvparam

Document the addition of the vnet_hdr HV parameter for KVM in the NEWS
file for 2.8.

Signed-off-by: Stratos Psomadakis <psomas@grnet.gr>
Reviewed-by: Guido Trotter <ultrotter@google.com>

10 years agoDocument the vnet_hdr HV parameter for KVM
Stratos Psomadakis [Mon, 29 Jul 2013 14:14:48 +0000 (17:14 +0300)]
Document the vnet_hdr HV parameter for KVM

Document the vnet_hdr HV parameter for KVM in the gnt-instance man page.

Signed-off-by: Stratos Psomadakis <psomas@grnet.gr>
Reviewed-by: Guido Trotter <ultrotter@google.com>

10 years agoFormat gnt-network info group output
Apollon Oikonomopoulos [Mon, 29 Jul 2013 12:50:00 +0000 (15:50 +0300)]
Format gnt-network info group output

gnt-network info currently displays the connected group output unformatted:

 # gnt-network info
 ...
   connected to node groups:
    ['med', 'bridged', 'staging']

This patch adds formatting to the node group output.

Signed-off-by: Apollon Oikonomopoulos <apoikos@gmail.com>
Reviewed-by: Guido Trotter <ultrotter@google.com>

10 years agoAdd 2.7.2 NEWS entry
Apollon Oikonomopoulos [Mon, 29 Jul 2013 12:49:41 +0000 (15:49 +0300)]
Add 2.7.2 NEWS entry

Prepare to list changes for 2.7.2.

Signed-off-by: Apollon Oikonomopoulos <apoikos@gmail.com>
Reviewed-by: Guido Trotter <ultrotter@google.com>

10 years agoPrepare NEWS for 2.8.0 beta2
Guido Trotter [Mon, 29 Jul 2013 11:52:56 +0000 (13:52 +0200)]
Prepare NEWS for 2.8.0 beta2

- Create stub "initial release" entry for beta1.
- Keep main 2.8 release information at top
- Add initial "since beta1" section

Signed-off-by: Guido Trotter <ultrotter@google.com>
Reviewed-by: Michele Tartara <mtartara@google.com>

10 years agoFix the downgrade function of cfgupgrade
Michele Tartara [Thu, 25 Jul 2013 17:17:18 +0000 (19:17 +0200)]
Fix the downgrade function of cfgupgrade

The downgrade was not correctly removing some of the UUIDs (namely, those
of disks and NICs).

Fixes Issue 510.

Signed-off-by: Michele Tartara <mtartara@google.com>
Reviewed-by: Klaus Aehlig <aehlig@google.com>

10 years agoIncrease maximum HTTP message size
Michele Tartara [Fri, 26 Jul 2013 10:12:35 +0000 (12:12 +0200)]
Increase maximum HTTP message size

After adding a new KVM parameter, RAPI queries related to KVM have become
too big, so we need to increase this.

Signed-off-by: Michele Tartara <mtartara@google.com>
Reviewed-by: Helga Velroyen <helgav@google.com>

10 years agoClarify hbal man page
Klaus Aehlig [Thu, 25 Jul 2013 17:05:45 +0000 (19:05 +0200)]
Clarify hbal man page

In the sense of hbal, and, in fact, all htools, drained nodes
are considered as offline. Also, the cluster score is not just
the sum of its component, but, e.g., the count of instances on
offline nodes is weighted stronger.

Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Michele Tartara <mtartara@google.com>

10 years agoProperly add the UUID to all the disks
Michele Tartara [Thu, 25 Jul 2013 14:00:17 +0000 (14:00 +0000)]
Properly add the UUID to all the disks

Starting from Ganeti 2.8 all the disks need to have a UUID.
A function for adding a UUID automatically to disks was present, but it didn't
consider disks with children (like DRBD).

The function is modified to work recursively.

Partially fixes Issue 510.

Signed-off-by: Michele Tartara <mtartara@google.com>
Reviewed-by: Helga Velroyen <helgav@google.com>

10 years agoMerge branch 'stable-2.7' into stable-2.8
Guido Trotter [Thu, 25 Jul 2013 12:38:24 +0000 (14:38 +0200)]
Merge branch 'stable-2.7' into stable-2.8

* stable-2.7:
  Update NEWS and version for Ganeti 2.7.1
  daemon-util: pass --oknodo at rotate_logs
  Add logrotate example
  daemon-util: provide rotate_logs and rotate_all_logs actions

Conflicts:
NEWS: trivial
configure.ac: trivial (keep 2.8 version)

Signed-off-by: Guido Trotter <ultrotter@google.com>
Reviewed-by: Michele Tartara <mtartara@google.com>

10 years agoUpdate NEWS and version for Ganeti 2.7.1 v2.7.1
Guido Trotter [Thu, 25 Jul 2013 10:36:54 +0000 (12:36 +0200)]
Update NEWS and version for Ganeti 2.7.1

Release 2.7.1 with a handful of minor fixes.

Signed-off-by: Guido Trotter <ultrotter@google.com>
Reviewed-by: Michele Tartara <mtartara@google.com>

10 years agoAdd hvparam to disable VNET_HDR on tap devices
Stratos Psomdakis [Wed, 24 Jul 2013 11:52:55 +0000 (14:52 +0300)]
Add hvparam to disable VNET_HDR on tap devices

IFF_VNET_HDR allows tap devices to extract a VIRTIO_NET_HDR from pkts
going through the tap iface, enabling segmentation offload for the
virtio nics.

Current (Debian) kernels (3.2 for Debian Squeeze backrports / Wheezy)
don't seem able to handle well VMs with 'broken' virtio ifaces (e.g.
FreeBSD virtio-net driver), using GSO. Broken or malicious virtio-net
drivers can cause the host kernel to flood the logs with WARNs(),
effectively live-locking the system and affecting its overall stabitlity
(eg DRBD disconnects).

The WARN() flooding has been fixed / ratelimited in Linus' tree, but it
hasn't been backported to -stable.

This patch adds the vnet_hdr hvparam for KVM, to select whether the tap
devices used for KVM virtio ifaces will get created with VNET_HDR
support (IFF_VNET_HDR), even when the underlying kernel supports it, in
order to avoid this issue.

Signed-off-by: Stratos Psomadakis <psomas@grnet.gr>
Reviewed-by: Guido Trotter <ultrotter@google.com>