This commits breaks down the "Input message" section of iallocator.rst into
two separate subsections: one detailing keys that are required in all
operation types; a second one detailing the "request" element, which is
different for each type of request.

Some other minor improvements are included as well:

  - update input example to version 2, and add the "nodegroups" and
    "enabled_hypervisors" top-level elements, and the "group" and
    "hypervisor" attributes for nodes and allocation request, respectively.

  - sort keys in the example dictionaries according to the order in earlier
    sections, for easy comparison of documentation with its examples.

Signed-off-by: Adeodato Simo <dato@google.com>
Reviewed-by: Iustin Pop <iustin@google.com>

This fixes a issue, where an stopped instances is reported as ERROR
in cluster verify if it lives on a offline node. As the instances is
down this shouldn't happen.

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

The design details can be seen in the design document
(doc/design-lu-generated-jobs.rst).

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

Signed-off-by: Adeodato Simo <dato@google.com>
Reviewed-by: Iustin Pop <iustin@google.com>

The request argument for multi-evacuate mode is "evac_nodes", not "nodes"
(the example later in the file has the correct name already).

Signed-off-by: Adeodato Simo <dato@google.com>
Reviewed-by: Iustin Pop <iustin@google.com>

Add some debug logging to detail why we don't run some steps.

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

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

I provided unittest to test the important pieces of the infrastructure.
The one remaining function (ResuriveEnsure) is not easy to unittest
but also not critical if it fails to operate correctly.

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

This allows Apache to display the directory in a nicer way.

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

OK, my previous small fix was not good. There is another issue: haddoc
(the documentation generator) needs to pass the same compiler options
(i.e. in our case, -DNO_CURL) to ghc. But in case of no curl, then it
shouldn't scan at all the RAPI library, as that is not used in our
builds.

Clearly, this is not a nice thing. So this patch changes from
including/excluding RAPI conditionally (in two places, the
ExtLoader.hs module and in hscan.hs), to always include RAPI, and
moves the curl/no curl login to RAPI itself, where it belongs.

Together with passing --optghc to haddock, this makes the builds
consistent both with and without RAPI. I also undo the removal of RAPI
from QC.hs.

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

hlint gives more suggestions, but some make the code (IMHO) harder to
read.

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

These are just changes from hlint suggestions. Still compiles and
passes unittests.

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

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

This patch adds a design document describing how jobs can be submitted
from within LUs.

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

Since we don't test for now the RAPI backend directly, we can simply
skip the import. Later we can make a conditional import if needed.

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>

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>

- Use m4_normalize to make single-line strings while removing
  unnecessary spaces
- Wrap lines longer than 80 characters

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

This documents the needed libraries for Haskell development. It also
fixes a tiny typo in devnotes.rst.

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

Based on Michael's suggestion, this patch partially reverts my
changes. The new directories are:

- doc/api/py
- doc/api/hs
- doc/coverage/py
- doc/coverage/hs

Basically the Python-specific output moves one level down (into py/)
compared to the original location, and the Haskell stuff goes into
hs/.

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

Also add mention about it being obsolete.

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

This should be more readable now. I wanted to even use the nicer
_SOURCES, but _SOURCES is special in Automake (again), so _SRCS it is.

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

Only one target wasn't ported over (the TAGS one), as hasktags is not
available easily in distributions, so it doesn't make sense to enable
it for all developers.

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

Since we do have errors currently, this is not enabled from the main
'make lint' rule. That will get cleaned up later.

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

This was a bit tricky, as the compilation from the top-dir changes the
paths in the .tix/.mix files.

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

Previously, Python api doc was under doc/api (which didn't match with
the target rule, apidoc). After this patch, we have the following:

- make py-apidoc generates Python api doc under doc/py-apidoc
- make hs-apidoc generates Haskell api doc under doc/hs-apidoc
- make apidoc does both (if hs-apidoc enabled at configure time)

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

Since we have two sets of coverage data, we make the naming
consistent:

- doc/py-coverage for Python coverage
- doc/hs-coverage for Haskell coverage

We also need to exclude some more dirs from the list of checked
directories.

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

We do it via per-target-binary .o/.hi files. There are a few other
options, in the end this was chosen as different binaries can have
different options/components (e.g. hail won't use RAPI, etc.) and that
the unittests need different compilation options.

We add a wildcard *.o/*.hi to CLEANFILES, so that it removes all
per-target build artifacts.

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

RAPI needs the curl module, which is a binding to the libcurl C
library. This is more complex (can be compiled either against gnutls
or openssl, etc.) and it's not of much help when the tools are used on
the cluster itself, we disable it by default.

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

Instead of bindir. We do this by adding a new allocdir variable, and
redirecting hail to it.

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>

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

This patch adds some (not very nice) rules for htools
compilation. Since automake only knows about some languages, and
doesn't have a real extension mechanism, I cheat and declare them as
"buildable scripts", instead of programs. Rationale:

automake support for dir_PROGRAMS is limited to a few languages; in
all cases, it assumes that translation from sources to binaries occurs
in two steps, compilation and linking; it's not easy to change its
assumption.

In our case, where “ghc --make” does both steps, it's hard to force it
into the automake model.

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

This patch adds an enable/disable option for htools (--enable-htools),
and associated tests for determining whether GHC (the compiler we use
for htools) and required libraries are present.

The method to do so is not very nice; usually, Haskell programs are
configured and compiled using cabal (http://www.haskell.org/cabal/

, a
tool similar to Python's setuptools)), but that doesn't suit itself to
nice integration with autoconf/automake, so we test for the modules
presence manually.

In the end, we set a few variables:

- GHC: the path to the ghc compiler
- HTOOLS_MODULES: command line option for ghc to select the wanted
  'parallel' module
- HTOOLS_NOCURL: set to -DNO_CURL if we don't want to enable curl (and
  thus RAPI) support in htools
- HTOOLS: set to 'yes' if we should compile/install the htools
  programs
- HTOOLS_APIDOC: set to yes if we should build/install the htools
  API documentation
- WANT_HTOOLS, WANT_HTOOLSAPIDOC: two automake conditionals for later
  use in Makefile.am

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

- Use constants
- Don't sleep if no delay is given
- Mark function as deprecated: it uses polling instead of waiting for changes
  (but the latter needs authentication); it can still be used
- Add unittests

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

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

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

Commit dd7f6776 added another call to BuildHooksEnv to provide
post-phase status variables. Since BuildHooksEnv also built the node
lists, that meant they have to be built twice. First a rather strict
check was used, but it turned out to be more tricky. Commit b423c513
had to remove the strict check again.

With this patch the function is split in two parts, one generating the
actual environment variables, and another part returning the node lists.
The former is called twice.

Unittests are updated.

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

cfgupgrade provides a --ignore-hostname option, however consequently it
spawns config.ConfigWriter, which refuses to handle a foreign
configuration file by default.

config.ConfigWriter provides a relative keyword argument,
accept_foreign, so we tie this to the value of options.ignore_hostname.

Signed-off-by: Apollon Oikonomopoulos <apollon@noc.grnet.gr>
Reviewed-by: Michael Hanselmann <hansmi@google.com>