Make the slots functionality more modular
As we will introduce another set of containers using the slots trickwe abstract away as much as possible to separate bases classes. Thechild classes then adapt them for their needs. This leads to less code...
Move _MakeLegacyNodeInfo to rpc
This function will also be needed for the iallocator refactor. Also thisfunction does not belong to cmdlib but rather to rpc. Lets move itthere.
Signed-off-by: René Nussbaumer <rn@google.com>Reviewed-by: Michael Hanselmann <hansmi@google.com>
Move _CalculateGroupIPolicy and _ComputeDiskSize
This makes it possible to better untangle the IAllocator code, whichwould otherwise lead to a recursive import structure as we need thosehelpers there as well.
Signed-off-by: René Nussbaumer <rn@google.com>...
Update documentation for "gnt-job list"
Commit f037e9d7 added command line flags to filter jobs, but thedocumentation was not updated. Also mention the ability to usemore complex filters.
Signed-off-by: Michael Hanselmann <hansmi@google.com>Reviewed-by: Iustin Pop <iustin@google.com>
Improve the prop_Alloc_sane test to detect mis-allocations
Currently, this just checks that a cluster cannot be rebalanced aftera single instance allocation. However, we can also test whether theallocation decision computed a correct new cluster score, by checking...
Add new test for checking multi-allocations
This test expands the "single-alloc-no-rebalance" by allocating a fewinstances on a small cluster, and ensuring that after we allocate allof them, either we can't rebalance or if we rebalance the scoreimprovement is very small....
Improve the `CanTieredAlloc' test
Currently, this test is very slow. Upon investigation, this is due tohow `tieredAlloc' works:
- tries to allocate one instance- if failed, shrink the instance by the "most failed" resource- restart
In this algorithm, if the "most failed" resource is e.g. memory, and...
Improve the `AllocPolicy' test
This test has a few deficiencies, which this patch addresses:
- using arbitrary 1 or 2 node count for allocation is obsolete, nowadays we need to use a number appropriate for the instance's disk template (and we should remove that parameter…)...
Fix arbitrary ConfigData object generation
The Cluster object, as it is defined right now, has many '[String]'members, which means that in a standard arbitrary generator these willbecome very big, which is the reason for the current slowness of thetest 'Config_serialisation'....
Remove the slow/fast tests functionality
Since the recent commits improved the speed of the two "slow" testgroups to regular test speed, we can remove this kludge and simplifysignificantly our test runner, yay!
Signed-off-by: Iustin Pop <iustin@google.com>...
Simplify a bit more the test harness
We can build the test groups directly in the `testSuite' helper,instead of doing it (much later) in the test harness.
Signed-off-by: Iustin Pop <iustin@google.com>Reviewed-by: Guido Trotter <ultrotter@google.com>
Merge branch 'devel-2.6' into submit
Signed-off-by: Iustin Pop <iustin@google.com>Reviewed-by: Agata Murawska <agatamurawska@google.com>
Merge branch 'stable-2.6' into devel-2.6
Fix bug in non-mirrored instance allocation
The function `allocateOnSingle' has a bug in the calculation of thecluster score used for deciding which of the many target nodes to usein placing the instance: it uses the original node list for the scorecalculation....
Fix deserialisation bug in ResultEntry
Found via the newly added unit-tests, which test most of theserialisation code in Query/Language (except for QueryResult, forwhich we already tests both sub-components separately).
Rework CLI modules and tests
While investigating how we could test the Daemon.hs module, I realisedthat we have a very, erm, sub-optimal situation:
- HTools/CLI.hs has a nice IO/pure separation testing in cmdline parsing, which allows some basic functionality to be tested, but...
Further hlint fixes
Commit 2cdaf22, “Re-enable standard hlint warnings”, got it almostright. The only problem is that (confusingly) the default set of hintsis not in HLint.Default, but in HLint.HLint (it includes Default andsome built-ins).
After changing the lint file to correctly include the defaults, we had...
Add query filter tests
These tests are node specific only because we don't have other querytypes implemented yet, but what they actually test is the variousfilter types.
The tests are trying to cover most filter functionality; missing fornow is proper checking for ContainsFilter and TrueFilter, the rest...
Replace manual arbitrary instances with genArbitrary
There are a few more that could be replaces, once we start usingappropriate (new)types.
Signed-off-by: Iustin Pop <iustin@google.com>Reviewed-by: René Nussbaumer <rn@google.com>
Add entire ConfigData serialisation tests
Using the recently-added genArbitrary, we can now implement Arbitraryinstances for even "huge" objects like Cluster, so let's use that toimplement entire ConfigData serialisation tests.
Note that, as we don't have yet proper types for some of the Params...
Add a small 'passTest' helper
This is symmetric to failTest, and allows us to use it in cases wherewe need to return a property.
While replacing 'property True' with it, I saw a case where we cansimplify the use and thus reworked that check.
Add some unittests for node queries
These new tests check that:
- no known fields return unknown- any unknown field returns unknown- the type of the fields is consistent between the getters and the field definition- the length of each result row corresponds with the number of fields...
Introduce a helper for simple Arbitrary instances
After getting really annoyed at yet another "<*> arbitrary", I thoughtthat we should be able to automate this, at least for types which aresimple enough and have already all the "prerequisites".
Hence the new genArbitrary function and its helpers, which can:...
Add unit test for serialisation of DiskLogicalId and Nodes
Since the DiskLogicalId type is manually serialised/deserialised (seeObjects.hs, `encodeDLid' and `decodeDLId'), let's add a test thatchecks that these are idempotent when combined.
Since we're at it, let's add the same test for Node serialisation,...
Add a test helper for simple JSON serialisation testing
While adding yet another JSON serialisation test, I realised that thiscan be trivially abstracted; hence this patch, replacing both simpleversions (readJSON . showJSON == id) and the standard version (with...
Add Instance serialisations tests
This is not perfect, as for many of the parameters we don't have goodArbitrary instances, but is better than nothing.
Add unittests for the BasicTypes module
This adds test properties for the various laws that the instances ofResult should follow; I could not find (offline) laws about `mappend',but otherwise I implemented all laws that I could find.
Note that we have to silence hlint warnings for the things we want to...
Fixup test suite names
The names were not in a proper hierarchy, leading to inconsistenciesabout what they were actually tested.
We change this by reproducing in the test names the relative hierarchywithin the Ganeti directory, leading to nicer test suite names (in...
Simplify property and test case names
Since we now have separate namespaces due to the multi-file split, wedon't need to keep the name of the module in the property names, as wedon't have so many potential conflicts anymore.
We remove the group prefix handling from TestHelper and simply do a...
Move Version.hs up from under HTools/
This is another module that is generic, and not htools-specific.
Two tiny makefile fixes
During the recent moves and renames, two things have slipped through,since I didn't run make check-local…
Remove QC.hs and replace it with an auto-generated file
In order to have correct code coverage results, we must somehow importall production modules into the test runner. Until now, this was donemanually (when we didn't forget) in QC.hs.
To improve the situation, we remove QC.hs and replace it with an...
Split the Makefile variable HS_LIB_SRCS
This currently holds both production files and test files; since wewant to treat these separately, let's split the variable in two, andintroduce another one for the old, inclusive one.
Split last two remaining tests from QC.hs
This makes QC.hs obsolete, but we will remove it in a later patch,once we automatically build a file with all the imports (to ensurecorrect coverage results).
Move JSON.hs and Compat.hs out from under HTools/
These two files are not htools-specific, so let's move them out of theHTools subdirectory/module hierarchy and directly under Ganeti.
Split most HTools test code into separate files
Except for Ganeti.HTools.JSON, which needs rename, we split all theother test suites into separate files.
We have to add another common test helper, due to import dependencies(sigh), but otherwise this split is pretty straightforward....
Split Luxi, Qlang, Ssconf and OpCodes tests
… from QC.hs into their own files, again mirroring the production codesource tree.
Split Rpc tests from QC
This required lots of other code moves, so I created it as astandalone patch.
Split one more module out of QC and add test helpers
This splits the confd/utils tests, and adds the TestCommon module forshared test code.
Split out Objects.hs from QC.hs
This is the first file split out from QC.hs - an easy one, since ithas just one test.
The patch changes the way we build hpc excludes, since now we'll havemany modules that need to be excluded, and hpc doesn't seem to be able...
Move the unittest helper to a new Test/Ganeti dir
This patch starts the move of the test haskell code from`htools/Ganeti/HTools/' to its more proper place of`htest/Test/Ganeti'.
Move generated htools test stubs to htest/
This moves the last (I think) htools-related bits out of test/ underhtest/.
Move shelltests from test/ to htest/shelltests/
This is another rather trivial patch, moving all the (htools)shelltests to their own directory.
Move htools test files from test/data/htools to htest/data
Following up on the program moves, we now move the test data files.
Move haskell test code to htest/
This is the first commit of a series that will attempt to cleanup thetest code organisation, which evolved somewhat organically from theinitial pure htools functionality.
The proposed organisation of the tree will be as follows:...
Fix gnt-debug iallocator
There was an issue with the recent ipolicy introduction which lead to abug in gnt-debug iallocator. It was not providing the spindle_use fieldand therefore it wont let you create a valid iallocator request.
Implement compilation of regexes at creation time
This means that the verification of the correctness of the regex isdone once, at the deserialisation/creation time, as in the Pythoncode. To do this, we have to change the FilterRegex type from an alias...
Rename the Qlang and Queryd modules
Per the new query module hierarchy, rename Qlang to Query/Language andQueryd to Query/Server. This way, all query-related functionality isnow "contained" in the Query/ directory.
Add filtering support in Query
This adds basic infrastructure for filtering (fully functional except,as usual, for runtime data), and then uses it for node queries.
Since the filtering exports regex matching as an externalfunctionality, we have to use a regex library. There are many flavours...
Add an Applicative instance for our Result type
This will allow us to run generic computations in an Applicativecontext, if a monadic one is not needed (or not applicable due toclass constraints).
Parameterize the Filter type
In preparation for introducing filtering functionality, we convert the'Filter' type from a '*' kind to a '* -> *' kind.
This allows us to define some general properties for the filter, andfor example introduce later an easy filter compilation, etc....
Implement QueryFields for Nodes
Since we have all the definitions already, we can easily enablethis. Manual testing shows no difference between the Python and theHaskell versions of node list-fields.
Merge branch 'devel-2.6'
Signed-off-by: Iustin Pop <iustin@google.com>Reviewed-by: Bernardo Dal Seno <bdalseno@google.com>
Putting the multiallocate pieces together
This is the final part:
Signed-off-by: René Nussbaumer <rn@google.com>Reviewed-by: Iustin Pop <iustin@google.com>
Adding allocList function
This function iterates over the tryMGAlloc function, updates the nodelist and instance list, and refeeds it recursively until no newinstances are left.
This allows us then to allocate multiple instances and see if they wouldfit together....
Adding new IAllocator mode to constants
Fix warnings/errors with newer pylint
To help developing Ganeti on newer distributions, let's try to fixpylint warnings/errors. I'm using pylint from current Debian wheezy:pylint 0.25.1, astng 0.23.1, common 0.58.0, and we have 3 things thatneeds fixing....
Signed-off-by: Guido Trotter <ultrotter@google.com>Reviewed-by: Iustin Pop <iustin@google.com>
Fix decorator uses which crash newer pylint
Pylint version:
pylint 0.25.1, astng 0.23.1, common 0.58.0
crashes when passing the fully-qualified decorator name with:
File "/usr/lib/pymodules/python2.7/pylint/checkers/base.py", line 161, in visit_function...
Instance autorepair design
This design describes a tool that will perform automatic repairs oninstances when they are detected to be unhealthy (living on offline ordrained nodes, at the moment). These repairs can be scheduledautomatically or requested as a one-off by a tool or person....
Add missing luxi query 'QueryFields'
This was missed; we add the definition and the de-serialisationsupport.
Add Query support for Nodes (no filtering, no RPC)
This is the initial support for Query2: basic infrastructure (exceptfiltering) and node query support (without RPC).
It implements all the fields (tests by comparison with list-fields onthe Python side), except that:...
Move parameter title definitions to constants
This fixes an old TODO for moving these; now they can be reused in theHaskell Query2 implementation.
One more SADT type - VType
This adds the VTYPE_* conversion to VType, used for QFT equivalencelater.
Extend convert-constants support for dicts
This enhances convert-constants to not flatten dicts completely, butalso generate a so-called association list for them.
This allows either direct use of the 'lookup' function, or (forperformance) conversion to Data.Map and optimised lookup later....
Expand THH with more functionality for parameters
This adds two related improvements to THH:
- for parameters, we declare a list with all their fields, so that Query2 can build the list of fields (e.g. for hvparams, or ndparams) automatically
- we declare a new type class for "DictObjects", i.e. objects which...
Add more node-related data types and functions
This is a simple type declaration for NodeRole, a NdParamObject typeclass, and a few related helper functions for nodes and node groups.
Fix typo in node role description
I'd rather not introduce this typo in the Haskell code too, and alsonot having them identical is not good.
Add type classes for common objects
These mirror the TaggableObject in Python, in the sense that we willbe able to define generic functions for querying such fields.
Stub query2 call integration into QueryD
This patch corrects the definitions in Qlang.hs to match what Pythonexpects on the wire; this means replacing some manual data typedefinitions with 'buildObject' so that we get serialisation (and fieldnames) for free, adding (manually) JSON instances for types which are...
Fix error reporting for bad Luxi arguments in QueryD
Currently, the query daemon would simply close the connection to theclient without issuing a response, if parsing the arguments failed;the error was just logged.
Since this is very ugly from the client's point of view, we change it...
Improve error reporting in our JSON conversions
Reporting things such as the following in our error messages(indentation added by me, not originally present, so it's even worse):
JSArray [JSArray [JSRational False (1 % 1),JSString (JSONString {fromJSString = "a"})],...
Cleanup query socket before attempting to bind
This patch adds a simple function to ensure a socket doesn't existbefore binding to it. Additionally, we cleanup the exports ofQueryd.hs.
Complete and enable the haskell QueryClusterInfo call
Since we have now access to all cluster parameters, we can "fill" theparameters, and can finally enable it in CLI/RAPI.
Expand Objects.hs definitions
This patch adds the missing parameters in the cluster/group objects,for now as simple maps (dictionaries), without type safety. Therationale for adding them as such is:
- we need something to enable query functionality- since we don't modify the values, we don't risk introducing bugs...
Add a fillDict function
This is similar to the Python version, objects.py:FillDict.
Fix qualified import of Data.Map in QC.hs
As opposed to all other places in the code, we imported this asData.Map, leaving to longer names. Let's make it similar to the restof the code (import qualified Data.Map as Map).
Enhance convert-constants list generation
While looking at the hypervisor types in Constants.hs, I saw that the'hyperTypes' list is using strings instead of names. This is due tothe fact that we require the entire elements in the list to beidentified (homogeneous lists), but the string "xen-pvm" is declared...
Add two confd library tests
These test that we encode requests correctly and decode good messages,that a message with a too old/new timestamp is never accepted, andthat a message signed with a different key is never accepted.
Reduce some more code duplication and split code
The Qlang module defines ResultStatus, but it was already defined inGaneti/Luxi.hs; let's remove the duplicate definition from there sincethe proper place is in the newer module.
Also, in order to ease testing, we split some confd functions into a...
Add test for checking Haskell/Python opcode equivalence
This is a very big hack for testing the equivalence of Python andHaskell opcode definitions. See the docstring for details; I'm notvery happy with the solution but it does the job.
An alternate option would be to launch the Python code when...
Create a custom type for disk indices
While (again) trying to test Python/Haskell encoding interoperability,I found another bug: the disk index is declared in Python asht.TPositiveInt, but in Haskell just as Int, so it can take negativevalues too. Clearly we can do better, so let's add a wrapper type that...
Handle better 'null' values in optional fields
While testing Haskell⇔Python interoperability for opcodeserialisation, I found this bug: the Haskell code doesn't treatoptional fields with 'null' values as missing, which the Python codedoes, leading to differences....
Add test case for OpCode list equivalence
For now, we only test that we don't define extra opcodes in theHaskell code; once we have parity, we can enable the (for now)commented-out lines that check the reverse.
Add infrastructure for reading Python command output
This will be used in the future to make tests/assertions forPython/Haskell codebase equivalence (e.g. checking that the sameopcodes are defined, etc.).
(Side-note: this might not work in VPATH builds; I'll see how it goes)...
QCHelper: add support for defining HUnit test cases
This will allow us to use exactly the same method as for defining andusing QuickCheck properties. The differentiation is based on name,which is not very nice but is the same method used intest-framework-th, for example (so we will be able to switch...
Rename tests for nicer names in test output
With the change to test-framework, where individual tests are listed,QCHelper tries to remove a "prop_$group_" prefix from the functionnames, for shorter names. However, some test groups (LUXI) and testfunctions are not uniform, leading to long names, so we should fix...
Switch Haskell test harness to test-framework
This patch replaces our home-grown, and quite limited, test runnerinfrastructure with test-framework(http://batterseapower.github.com/test-framework/). The rationale fordoing so is as follows:
- we will need to add support for HUnit tests, so either we add more...
Simplify the Luxi client-server test
Since now we have the recvMsgExt, we can remove the handleEOFfunction.
Fix bug in the Luxi client-server test
We didn't close the server correctly, so for test sizes bigger thanthe max-open-files ulimit, the tests were failing in accept().
TH: one style fix and more docstrings
We were missing many docstrings in THH.hs, so let's add at least someof them, and fix some unquoted '/'. Additionally one style change hasbeen done.
OpCodes: build and export a list of all opcodes
This can be used for cross-checking with the Python code forconsistency on defined opcodes.
TH: Abstract function for computing constructor names
We'll need this in another place shortly, so let's abstract it and addproper verification of whether we were passed a type name correctly;the previous version would have failed with a pattern match failure,...
Enable tags query over the query socket
This patch adds the tags field to the objects which were missing it(all except Cluster), implements handling the LuxiCall QueryTags, andthen enables the use of the query socket in cli.ListTags, used by allcommands, and in the RAPI client....
Switch RAPI to use same tags listing as CLI
Currently, RAPI uses queries (as in QueryNode with field tags) forgetting the tags, whereas the CLI uses the QueryTags Luxicommand. Since this discrepancy is not good, let's switch RAPI to usethe same method as the CLI....
Config.hs: Add a function to lookup group
Like in the python code, this is a bit more complex since groups areindexed by UUID, so we must fallback to lookup via the name (slow; butthe number of groups should be small).
Add a Functor instance for our Result type
This will allow us to use the simpler applicative form (<$>, <*>)instead of monadic liftM, liftM2, etc.
Change the Luxi tags kind from String to a custom type
This will allow safer code when we implement the tags query.
Remove obsolete QrViaLuxi type
The actual query definitions are now in Qlang.hs, so let's use theItemType from there instead of luxi-defined type (which is alsoincomplete).