GitLab

The test framework we use sets PYTHONPATH for tests so that
the python module under test is used. Unfortunately, the paths
in PYTHONPATH are relative paths. So we need to expand them for
this test. Here we just update our fragile work around, but the
only way the test is called is, well, in testing, so we can build
it for our test environment.
Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Helga Velroyen <helgav@google.com>

* stable-2.9
  Make ganeti-cleaner switch to a save working directory
Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Petr Pudlak <pudlak@google.com>

For some reason, find does stat the current working directory,
which, when invoked via cron, may or may not be readable by the
process (issue 880). Therefore, first change to a directory that
is definitely accessible to the process; the directory it logs
into is such a safe place.
Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Petr Pudlak <pudlak@google.com>

For a cluster that existed for a long time, the archive of
old jobs might be big. Additionally, this data is neither
necessary for running new jobs, nor is it touched in the
upgrade procedure. Hence it is safe to be excluded from
the backup created on Ganeti version upgrades, thus obtaining
a smaller backup.
Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Petr Pudlak <pudlak@google.com>

This patch exposes the early_release argument of the jobs resource on
the RAPI python wrapper, making it possible to use early_release when
replacing instance disks.
Signed-off-by: Fabian Geisberger <fabi@google.com>
Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Klaus Aehlig <aehlig@google.com>

Make gnt-cluster upgrade honor the --with-backup-dir configure option.
In this way, at configure time, it can be decided in which directory
the configuration backup tar balls are put.
Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Hrvoje Ribicic <riba@google.com>

During Ganeti version upgrades, Ganeti creates a backup of
the full configuration directory. Add a new configure option
to decide in which directory the backup tar ball is to be put.
Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Hrvoje Ribicic <riba@google.com>

In case of non lvm-based disk (i.e. drbd, plain disk template)
snapshot is currently not supported. The RPC will fail reporting the
disk's logical_id (and not unique_id since only BlockDev has one)
and its dev_type.
Signed-off-by: Dimitris Aragiorgis <dimara@grnet.gr>
Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Klaus Aehlig <aehlig@google.com>

Let RAPI raise HttpNotFound() in case the item requested is not
found, if luxid serves the query (split query mechanism).
Signed-off-by: Dimitris Aragiorgis <dimara@grnet.gr>
Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Klaus Aehlig <aehlig@google.com>

Up until now, in case the KVM hypervisor used chroot or pool/user
security model, disk hot-remove was not allowed. This should not be
the case since only during hot-add there are permission constrains.

Fix that.
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 a piece of convoluted code in the
implementation of instance queries. There was an
unecessarily nested loop, which in case of ongoing
migrations, could throw a key error. This fixes
issue 875.
Signed-off-by: Helga Velroyen <helgav@google.com>
Reviewed-by: Klaus Aehlig <aehlig@google.com>

* stable-2.9
  Push return out of genericResult
  Make htools tolerate missing "spfree" on luxi
Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Petr Pudlak <pudlak@google.com>

...and merge with '<-' to get a pure computation
outside the monad.
Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Petr Pudlak <pudlak@google.com>

If a cluster allows sharedfile as only disk template, the
amount of free spindles might not be available. This is
perfectly normal, hence the luxi backend of htools should
handle this case gracefully and just report 0 available spindles.
Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Hrvoje Ribicic <riba@google.com>

* stable-2.9
  Move the design for query splitting to the implemented list

Conflicts:
	doc/design-draft.rst
Resolution:
  Fix the missing .rst extension in 2.10.
Signed-off-by: Petr Pudlak <pudlak@google.com>
Reviewed-by: Klaus Aehlig <aehlig@google.com>

Signed-off-by: Petr Pudlak <pudlak@google.com>
Reviewed-by: Klaus Aehlig <aehlig@google.com>

* stable-2.9
  Add tests for DRBD setups with empty first resource

* stable-2.8
  DRBD parser: consume initial empty resource lines
Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Petr Pudlak <pudlak@google.com>

Add a DRBD 8.4 configuration with the first resource empty, together
with the relevant Haskell and Python tests.
Signed-off-by: Apollon Oikonomopoulos <apoikos@gmail.com>
Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Klaus Aehlig <aehlig@google.com>

* stable-2.8
  DRBD parser: consume initial empty resource lines
Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Petr Pudlak <pudlak@google.com>

When a DRBD resource is taken down, its minor number is freed and its
place in /proc/drbd is taken by an empty line. Currently the parser is
able to consume any empty lines appearing inside the resource stanza,
but cannot handle the first resource being empty.

We fix this by consuming all EOL characters appearing before the device
number. A test for this case should also be added.

This fixes issue #869.
Signed-off-by: Apollon Oikonomopoulos <apoikos@gmail.com>
Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Klaus Aehlig <aehlig@google.com>

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

...and schedule release date. This is a bug-fix release.
Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Helga Velroyen <helgav@google.com>

Make network_vlan OpParam of OpNetworkConnect default to the empty
string and thus make it optional (see defaultField definition). This
way we keep backwards compatibility with old RAPI.

Additionally bypass a sphinx warning related to the generated html
(i.e. Inline literal start-string without end-string) in case an
empty string is used as default value.
Signed-off-by: Dimitris Aragiorgis <dimara@grnet.gr>
Reviewed-by: Hrvoje Ribicic <riba@google.com>

...and warn that this option usually leads to an overestimation
of the cluster capacity.
Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Hrvoje Ribicic <riba@google.com>

...by verifying that instances are found, even in the
single-group scenario where some nodes are not N+1 happy.
Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Hrvoje Ribicic <riba@google.com>

Make hspace honor the option --accept-existing-errors. If
this option is given, hspace will try to allocate on all
N+1 happy nodes even though this might lead to overfull
groups.
Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Hrvoje Ribicic <riba@google.com>

This option will be used to tell hspace to ignore
existing N+1 violations and allocate on all non-affected
nodes as usual.
Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Hrvoje Ribicic <riba@google.com>

...in the hspace man page. Also document the default behavior
of hspace if this option is not given.
Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Hrvoje Ribicic <riba@google.com>

Verify that the results of allocation do not depend on the
presence of an overfull group, if the option --independent-groups
is given.
Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Hrvoje Ribicic <riba@google.com>

If this option is given, relax the check for a previous
not N+1 happy cluster to ignore the affected group, but
still consider allocations in other groups.
Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Hrvoje Ribicic <riba@google.com>

This option will make hspace consider node groups as
independent; in particular it will tell hspace not
to consider an N+1 violation in one group a reason
not to allocate in a different group.
Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Hrvoje Ribicic <riba@google.com>

In this way, algorithms doing allocation, like hspace,
can conveniently specify that they want to ignore this
group.
Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Hrvoje Ribicic <riba@google.com>

...when verbosity level is high enough.
Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Hrvoje Ribicic <riba@google.com>

The strict check if a tuple is indeed a tuple in the result validation
system is too strict for the current assumptions in the system. This
patch permits lists to be used when tuples are expected, provided they
still satisfy the type constraints.
Signed-off-by: Hrvoje Ribicic <riba@google.com>
Reviewed-by: Petr Pudlak <pudlak@google.com>

Signed-off-by: Hrvoje Ribicic <riba@google.com>
Reviewed-by: Klaus Aehlig <aehlig@google.com>

The tuple validation for opcode results failed to assert if a result
was a tuple before attempting to check its length and types. This
patch adds the missing "Is it a tuple?" check.
Signed-off-by: Hrvoje Ribicic <riba@google.com>
Reviewed-by: Klaus Aehlig <aehlig@google.com>

The vlan nicparam is used in openvswitch mode. This patch
exports it as VLAN environment variable so that NIC
configuration scripts can later use it.
Signed-off-by: Dimitris Aragiorgis <dimara@grnet.gr>
Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Klaus Aehlig <aehlig@google.com>

Cherry-picked from 747ee762.
Signed-off-by: Dimitris Aragiorgis <dimara@grnet.gr>
Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Klaus Aehlig <aehlig@google.com>

Until now gnt-network connect expected the network,
mode, link, and optionally groups as positional arguments.

To be able to pass vlan info too, use the `--nic-parameters`
option, just like gnt-cluster client to pass the desired
mode, link, and vlan. Missing values will be filled with the
default ones.

Do not support old argument passing since mode and link should
be arguments with variable length and this violates bash completion
generation rules.

Update QA to pass the new option instead of deprecated arguments.
Signed-off-by: Dimitris Aragiorgis <dimara@grnet.gr>
Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Klaus Aehlig <aehlig@google.com>

Let query mechanism (both python and haskell) be aware of the
vlan info included in a network's netparams.
Signed-off-by: Dimitris Aragiorgis <dimara@grnet.gr>
Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Klaus Aehlig <aehlig@google.com>

Upon LUNetworkDisconnect() export GROUP_NETWORK_MODE and
GROUP_NETWORK_LINK along with GROUP_NAME.

Export GROUP_NETWORK_VLAN too.
Signed-off-by: Dimitris Aragiorgis <dimara@grnet.gr>
Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Klaus Aehlig <aehlig@google.com>