GitLab

The data collector for /proc/diskstats was undocumented. Add an entry for it in
the mon-collector man page.

Also, remove some end of line whitespaces from the file.
Signed-off-by: Michele Tartara <mtartara@google.com>
Reviewed-by: Klaus Aehlig <aehlig@google.com>

* stable-2.8
  Fix wrong release date in the NEWS file
  Version bump for 2.8.0
  Add daemon split design doc

* stable-2.7
  Version bump for 2.7.2

Conflicts:
	Makefile.am
	NEWS
	configure.ac
	doc/design-draft.rst
configure.ac: ignore version change in stable-2.8
rest: take both additions
Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Thomas Thrainer <thomasth@google.com>

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

Also, update the NEWS file accordingly.
Signed-off-by: Michele Tartara <mtartara@google.com>
Reviewed-by: Klaus Aehlig <aehlig@google.com>

This describes the future planned structure of Ganeti daemons.
Signed-off-by: Michele Tartara <mtartara@google.com>
Reviewed-by: Guido Trotter <ultrotter@google.com>

The data collectors should be able to provide as much information as possible
even when the system is badly degraded. This patch modifies the instance status
collector for xen so that it can keep providing as much data as possible, even
when some of the queries it performs fail, by removing exitIfBad calls and
substituting them with logging and returning empty fields instead.
Signed-off-by: Michele Tartara <mtartara@google.com>
Reviewed-by: Klaus Aehlig <aehlig@google.com>

All data traffic usually goes over the secondary network, but
gnt-instance move didn't. This patch corrects this problem by using the
target nodes secondary IP as move target.
Signed-off-by: Thomas Thrainer <thomasth@google.com>
Reviewed-by: Guido Trotter <ultrotter@google.com>
(cherry picked from commit f56ab6d1)

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

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>

* stable-2.8
  Add additional tests for utils.Retry
  Make retry tests independent of actual time
  Fix corner-case in handling of remaining retry time
  Perform proper cleanup on termination of Haskell daemons
Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Michele Tartara <mtartara@google.com>

Signed-off-by: Robin H. Johnson <robbat2@gentoo.org>

The patch was released by the author into the public domain.
Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Klaus Aehlig <aehlig@google.com>

Linting is now verified against the lint tools version numbers as listed in this
patch.

These version are corresponding to the ones that can be found in the
repositories of Ubuntu 13.04.
Signed-off-by: Michele Tartara <mtartara@google.com>
Reviewed-by: Helga Velroyen <helgav@google.com>

Fix all instances of pep8's error: "E128 continuation line under-indented for
visual indent".
Signed-off-by: Michele Tartara <mtartara@google.com>
Reviewed-by: Guido Trotter <ultrotter@google.com>

More recent versions of pep8 check for the indentation level of closing
parenthesis in a way that is incompatible with the way we use in our code. Our
style is still pep8 compliant, and was re-allowed in even more recent versions
of the tool, but still we have to disable the check, for the time being.
Signed-off-by: Michele Tartara <mtartara@google.com>
Reviewed-by: Guido Trotter <ultrotter@google.com>

Mark a string representing a "sed" command including a regular expression as a
raw string, so that the entire escaping is performed by "sed" itself.
Signed-off-by: Michele Tartara <mtartara@google.com>
Reviewed-by: Klaus Aehlig <aehlig@google.com>

Newer pylint versions complain that QaConfig is not a proper container because
it does not implement the __len__(), __delitem__() and __setitem__() methods.
This commit implements them.
Signed-off-by: Michele Tartara <mtartara@google.com>
Reviewed-by: Klaus Aehlig <aehlig@google.com>

Newer pylint versions complain that QmpMessage is not a proper container because
it does not implement the __len__() and __delitem__() methods. This commit
implements them.

Also, tests are added.
Signed-off-by: Michele Tartara <mtartara@google.com>
Reviewed-by: Helga Velroyen <helgav@google.com>

In order to check for the correctness of the accesses to fields of objects,
pylint uses astng, but this is known to fail on some dynamically generated
classes. Still, newer pylint versions are stricter at performing this check, and
therefore fail while testing our codebase. Disable the checking of a few
selected classes that incur in this problem.
Signed-off-by: Michele Tartara <mtartara@google.com>
Reviewed-by: Thomas Thrainer <thomasth@google.com>

Recent versions of pylint require the first parameter of a metaclass to be
"cls". Older versions want "mcs". There's no way to support both by changing the
code, so the configuration file of pylint is changed instead, to have the
current codebase work with both versions.
Signed-off-by: Michele Tartara <mtartara@google.com>
Reviewed-by: Klaus Aehlig <aehlig@google.com>

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>

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>

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>

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>

Generating ASCII characters via a supported but not official escape sequence
leads to a "Anomalous backslash in string" error in newer pylint versions. This
patch fixes the issue.
Signed-off-by: Michele Tartara <mtartara@google.com>
Reviewed-by: Klaus Aehlig <aehlig@google.com>

Between the last two merges, nothing happened on stable-2.9. The
only change on stable-2.8 that touched lib/cmdlib/instance.py is
ff34fb97

. So make sure, the net change since the last merge to
this file is that of the said commit. This also fixes the doclint
errors.
Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Michele Tartara <mtartara@google.com>

* stable-2.8
  Add a default to file-driver when unspecified over RAPI
  Remove typo from RAPI documentation
  Describe the files in doc/users in NEWS
  Update the upgrade procedure

* stable-2.7
  Fix incorrect manpage reference to htools

Conflicts:
	NEWS
	configure.ac
	lib/cmdlib/instance.py
	man/hroller.rst

All trivial, except for man/hrooler.rst, where the
fix in the section number was manually applied to the
new text.
Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Thomas Thrainer <thomasth@google.com>

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>

* stable-2.7
  Fix incorrect manpage reference to htools
Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Michele Tartara <mtartara@google.com>

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>

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>

"be used instead." was repeated twice.
Signed-off-by: Michele Tartara <mtartara@google.com>
Reviewed-by: Klaus Aehlig <aehlig@google.com>

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>

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>

* stable-2.8
  Make tools/users-setup explain its actions first
  Generate tools/users-setup from descriptions
  Provide ganeti user and group data in easy-to-parse form
  Remove dsahostkeypub during config downgrade
  Mark the DSA host pubkey as optional
  Fix documentation for gnt-node evacuate -p option

Conflicts:
	Makefile.am (trivial)
	tools/cfgupgrade (removed downgrade code to 2.7)
Signed-off-by: Thomas Thrainer <thomasth@google.com>
Reviewed-by: Klaus Aehlig <aehlig@google.com>

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>

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>

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>

Bridging does not work correctly in net-common. The parameters passed
by Ganeti are put in /var/run/ganeti/xen-hypervisor/nic/$domname/$devid
and read again from there before running the setup functions according
to the $MODE which is read from the file.

When using MODE=bridged, the script expects to get the name of the
bridge passed as $BRIDGE, but is actually stored and passed as $LINK.
This patch fixes the issue.
Signed-off-by: Sebastian Gebhard <sege@fs.ei.tum.de>
Signed-off-by: Klaus Aehlig <aehlig@google.com>
Reviewed-by: Klaus Aehlig <aehlig@google.com>

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>

Commit a9542a4f

 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>