.. contents:: :depth: 4
-Objective
-=========
-
-Background
-==========
-
-Overview
-========
-
-Detailed design
-===============
-
As for 2.1 we divide the 2.2 design into three areas:
- core changes, which affect the master daemon/job queue/locking or
all/most logical units
- logical unit/feature changes
-- external interface changes (eg. command line, os api, hooks, ...)
+- external interface changes (e.g. command line, OS API, hooks, ...)
+
Core changes
-------------
+============
Master Daemon Scaling improvements
-~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+----------------------------------
Current state and shortcomings
-++++++++++++++++++++++++++++++
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Currently the Ganeti master daemon is based on four sets of threads:
Also, with the current architecture, masterd suffers from quite a few
scalability issues:
-- Since the 16 client worker threads handle one connection each, it's
- very easy to exhaust them, by just connecting to masterd 16 times and
- not sending any data. While we could perhaps make those pools
- resizable, increasing the number of threads won't help with lock
- contention.
-- Some luxi operations (in particular REQ_WAIT_FOR_JOB_CHANGE) make the
- relevant client thread block on its job for a relatively long time.
- This makes it easier to exhaust the 16 client threads.
-- The job queue lock is quite heavily contended, and certain easily
- reproducible workloads show that's it's very easy to put masterd in
- trouble: for example running ~15 background instance reinstall jobs,
- results in a master daemon that, even without having finished the
- client worker threads, can't answer simple job list requests, or
- submit more jobs.
+Core daemon connection handling
++++++++++++++++++++++++++++++++
+
+Since the 16 client worker threads handle one connection each, it's very
+easy to exhaust them, by just connecting to masterd 16 times and not
+sending any data. While we could perhaps make those pools resizable,
+increasing the number of threads won't help with lock contention nor
+with better handling long running operations making sure the client is
+informed that everything is proceeding, and doesn't need to time out.
+
+Wait for job change
++++++++++++++++++++
+
+The REQ_WAIT_FOR_JOB_CHANGE luxi operation makes the relevant client
+thread block on its job for a relative long time. This is another easy
+way to exhaust the 16 client threads, and a place where clients often
+time out, moreover this operation is negative for the job queue lock
+contention (see below).
+
+Job Queue lock
+++++++++++++++
+
+The job queue lock is quite heavily contended, and certain easily
+reproducible workloads show that's it's very easy to put masterd in
+trouble: for example running ~15 background instance reinstall jobs,
+results in a master daemon that, even without having finished the
+client worker threads, can't answer simple job list requests, or
+submit more jobs.
+
+Currently the job queue lock is an exclusive non-fair lock insulating
+the following job queue methods (called by the client workers).
+
+ - AddNode
+ - RemoveNode
+ - SubmitJob
+ - SubmitManyJobs
+ - WaitForJobChanges
+ - CancelJob
+ - ArchiveJob
+ - AutoArchiveJobs
+ - QueryJobs
+ - Shutdown
+
+Moreover the job queue lock is acquired outside of the job queue in two
+other classes:
+
+ - jqueue._JobQueueWorker (in RunTask) before executing the opcode, after
+ finishing its executing and when handling an exception.
+ - jqueue._OpExecCallbacks (in NotifyStart and Feedback) when the
+ processor (mcpu.Processor) is about to start working on the opcode
+ (after acquiring the necessary locks) and when any data is sent back
+ via the feedback function.
+
+Of those the major critical points are:
+
+ - Submit[Many]Job, QueryJobs, WaitForJobChanges, which can easily slow
+ down and block client threads up to making the respective clients
+ time out.
+ - The code paths in NotifyStart, Feedback, and RunTask, which slow
+ down job processing between clients and otherwise non-related jobs.
+
+To increase the pain:
+
+ - WaitForJobChanges is a bad offender because it's implemented with a
+ notified condition which awakes waiting threads, who then try to
+ acquire the global lock again
+ - Many should-be-fast code paths are slowed down by replicating the
+ change to remote nodes, and thus waiting, with the lock held, on
+ remote rpcs to complete (starting, finishing, and submitting jobs)
Proposed changes
-++++++++++++++++
+~~~~~~~~~~~~~~~~
In order to be able to interact with the master daemon even when it's
under heavy load, and to make it simpler to add core functionality
understand, debug, and scale.
Connection handling
-^^^^^^^^^^^^^^^^^^^
++++++++++++++++++++
We'll move the main thread of ganeti-masterd to asyncore, so that it can
share the mainloop code with all other Ganeti daemons. Then all luxi
thread on the socket.
Wait for job change
-^^^^^^^^^^^^^^^^^^^
++++++++++++++++++++
The REQ_WAIT_FOR_JOB_CHANGE luxi request is changed to be
subscription-based, so that the executing thread doesn't have to be
Other features to look at, when implementing this code are:
- - Possibility not to need the job lock to know which updates to push.
+ - Possibility not to need the job lock to know which updates to push:
+ if the thread producing the data pushed a copy of the update for the
+ waiting clients, the thread sending it won't need to acquire the
+ lock again to fetch the actual data.
- Possibility to signal clients about to time out, when no update has
been received, not to despair and to keep waiting (luxi level
keepalive).
them at a maximum rate (lower priority).
Job Queue lock
-^^^^^^^^^^^^^^
-
-Our tests show that the job queue lock is a point of high contention.
-We'll try to decrease its contention, either by more granular locking,
-the use of shared/exclusive locks, or reducing the size of the critical
-sections. This section of the design should be updated with the proposed
-changes for the 2.2 release, with regards to the job queue.
-
-Remote procedure call timeouts
-~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
-
-Current state and shortcomings
-++++++++++++++++++++++++++++++
-
-The current RPC protocol used by Ganeti is based on HTTP. Every request
-consists of an HTTP PUT request (e.g. ``PUT /hooks_runner HTTP/1.0``)
-and doesn't return until the function called has returned. Parameters
-and return values are encoded using JSON.
-
-On the server side, ``ganeti-noded`` handles every incoming connection
-in a separate process by forking just after accepting the connection.
-This process exits after sending the response.
-
-There is one major problem with this design: Timeouts can not be used on
-a per-request basis. Neither client or server know how long it will
-take. Even if we might be able to group requests into different
-categories (e.g. fast and slow), this is not reliable.
-
-If a node has an issue or the network connection fails while a request
-is being handled, the master daemon can wait for a long time for the
-connection to time out (e.g. due to the operating system's underlying
-TCP keep-alive packets or timeouts). While the settings for keep-alive
-packets can be changed using Linux-specific socket options, we prefer to
-use application-level timeouts because these cover both machine down and
-unresponsive node daemon cases.
-
-Proposed changes
-++++++++++++++++
-
-RPC glossary
-^^^^^^^^^^^^
-
-Function call ID
- Unique identifier returned by ``ganeti-noded`` after invoking a
- function.
-Function process
- Process started by ``ganeti-noded`` to call actual (backend) function.
-
-Protocol
-^^^^^^^^
-
-Initially we chose HTTP as our RPC protocol because there were existing
-libraries, which, unfortunately, turned out to miss important features
-(such as SSL certificate authentication) and we had to write our own.
-
-This proposal can easily be implemented using HTTP, though it would
-likely be more efficient and less complicated to use the LUXI protocol
-already used to communicate between client tools and the Ganeti master
-daemon. Switching to another protocol can occur at a later point. This
-proposal should be implemented using HTTP as its underlying protocol.
-
-The LUXI protocol currently contains two functions, ``WaitForJobChange``
-and ``AutoArchiveJobs``, which can take a longer time. They both support
-a parameter to specify the timeout. This timeout is usually chosen as
-roughly half of the socket timeout, guaranteeing a response before the
-socket times out. After the specified amount of time,
-``AutoArchiveJobs`` returns and reports the number of archived jobs.
-``WaitForJobChange`` returns and reports a timeout. In both cases, the
-functions can be called again.
-
-A similar model can be used for the inter-node RPC protocol. In some
-sense, the node daemon will implement a light variant of *"node daemon
-jobs"*. When the function call is sent, it specifies an initial timeout.
-If the function didn't finish within this timeout, a response is sent
-with a unique identifier, the function call ID. The client can then
-choose to wait for the function to finish again with a timeout.
-Inter-node RPC calls would no longer be blocking indefinitely and there
-would be an implicit ping-mechanism.
-
-Request handling
-^^^^^^^^^^^^^^^^
-
-To support the protocol changes described above, the way the node daemon
-handles request will have to change. Instead of forking and handling
-every connection in a separate process, there should be one child
-process per function call and the master process will handle the
-communication with clients and the function processes using asynchronous
-I/O.
-
-Function processes communicate with the parent process via stdio and
-possibly their exit status. Every function process has a unique
-identifier, though it shouldn't be the process ID only (PIDs can be
-recycled and are prone to race conditions for this use case). The
-proposed format is ``${ppid}:${cpid}:${time}:${random}``, where ``ppid``
-is the ``ganeti-noded`` PID, ``cpid`` the child's PID, ``time`` the
-current Unix timestamp with decimal places and ``random`` at least 16
-random bits.
-
-The following operations will be supported:
-
-``StartFunction(fn_name, fn_args, timeout)``
- Starts a function specified by ``fn_name`` with arguments in
- ``fn_args`` and waits up to ``timeout`` seconds for the function
- to finish. Fire-and-forget calls can be made by specifying a timeout
- of 0 seconds (e.g. for powercycling the node). Returns three values:
- function call ID (if not finished), whether function finished (or
- timeout) and the function's return value.
-``WaitForFunction(fnc_id, timeout)``
- Waits up to ``timeout`` seconds for function call to finish. Return
- value same as ``StartFunction``.
-
-In the future, ``StartFunction`` could support an additional parameter
-to specify after how long the function process should be aborted.
-
-Simplified timing diagram::
-
- Master daemon Node daemon Function process
- |
- Call function
- (timeout 10s) -----> Parse request and fork for ----> Start function
- calling actual function, then |
- wait up to 10s for function to |
- finish |
- | |
- ... ...
- | |
- Examine return <---- | |
- value and wait |
- again -------------> Wait another 10s for function |
- | |
- ... ...
- | |
- Examine return <---- | |
- value and wait |
- again -------------> Wait another 10s for function |
- | |
- ... ...
- | |
- | Function ends,
- Get return value and forward <-- process exits
- Process return <---- it to caller
- value and continue
- |
-
-.. TODO: Convert diagram above to graphviz/dot graphic
-
-On process termination (e.g. after having been sent a ``SIGTERM`` or
-``SIGINT`` signal), ``ganeti-noded`` should send ``SIGTERM`` to all
-function processes and wait for all of them to terminate.
+++++++++++++++
+
+In order to decrease the job queue lock contention, we will change the
+code paths in the following ways, initially:
+
+ - A per-job lock will be introduced. All operations affecting only one
+ job (for example feedback, starting/finishing notifications,
+ subscribing to or watching a job) will only require the job lock.
+ This should be a leaf lock, but if a situation arises in which it
+ must be acquired together with the global job queue lock the global
+ one must always be acquired last (for the global section).
+ - The locks will be converted to a sharedlock. Any read-only operation
+ will be able to proceed in parallel.
+ - During remote update (which happens already per-job) we'll drop the
+ job lock level to shared mode, so that activities reading the lock
+ (for example job change notifications or QueryJobs calls) will be
+ able to proceed in parallel.
+ - The wait for job changes improvements proposed above will be
+ implemented.
+
+In the future other improvements may include splitting off some of the
+work (eg replication of a job to remote nodes) to a separate thread pool
+or asynchronous thread, not tied with the code path for answering client
+requests or the one executing the "real" work. This can be discussed
+again after we used the more granular job queue in production and tested
+its benefits.
Inter-cluster instance moves
-~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+----------------------------
Current state and shortcomings
-++++++++++++++++++++++++++++++
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
With the current design of Ganeti, moving whole instances between
different clusters involves a lot of manual work. There are several ways
this process in Ganeti 2.2.
Proposed changes
-++++++++++++++++
+~~~~~~~~~~~~~~~~
Authorization, Authentication and Security
-^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+++++++++++++++++++++++++++++++++++++++++++
Until now, each Ganeti cluster was a self-contained entity and wouldn't
talk to other Ganeti clusters. Nodes within clusters only had to trust
not know the secret, it can't forge the certificates or redirect the
data. Unless disabled by a new cluster parameter, verifying the HMAC
signatures must be mandatory. The HMAC signature for X509 certificates
-will be prepended to the certificate similar to an RFC822 header and
+will be prepended to the certificate similar to an :rfc:`822` header and
only covers the certificate (from ``-----BEGIN CERTIFICATE-----`` to
``-----END CERTIFICATE-----``). The header name will be
``X-Ganeti-Signature`` and its value will have the format
certificate while providing a client certificate to the server.
Copying data
-^^^^^^^^^^^^
+++++++++++++
To simplify the implementation, we decided to operate at a block-device
level only, allowing us to easily support non-DRBD instance moves.
Intra-cluster instance moves will re-use the existing export and import
scripts supplied by instance OS definitions. Unlike simply copying the
-raw data, this allows to use filesystem-specific utilities to dump only
-used parts of the disk and to exclude certain disks from the move.
+raw data, this allows one to use filesystem-specific utilities to dump
+only used parts of the disk and to exclude certain disks from the move.
Compression should be used to further reduce the amount of data
transferred.
directly, where it'll be written to the new block device directly again.
Workflow
-^^^^^^^^
+++++++++
#. Third party tells source cluster to shut down instance, asks for the
instance specification and for the public part of an encryption key
- Instance information can already be retrieved using an existing API
- (``OpQueryInstanceData``).
+ (``OpInstanceQueryData``).
- An RSA encryption key and a corresponding self-signed X509
certificate is generated using the "openssl" command. This key will
be used to encrypt the data sent to the destination cluster.
instance move with the key received from the source cluster and
receives the public part of the destination's encryption key
- - The current API to create instances (``OpCreateInstance``) will be
+ - The current API to create instances (``OpInstanceCreate``) will be
extended to support an import from a remote cluster.
- A valid, unexpired X509 certificate signed with the destination
cluster's secret will be required. By verifying the signature, we
- The destination cluster generates another key and certificate,
signs and sends it to the third party, who will have to pass it to
- the API for exporting an instance (``OpExportInstance``). This
+ the API for exporting an instance (``OpBackupExport``). This
certificate is used to ensure we're sending the disk data to the
correct destination cluster.
- Once a disk can be imported, the API sends the destination
together with all necessary information to source cluster and tells
it to start the move
- - The existing API for exporting instances (``OpExportInstance``)
+ - The existing API for exporting instances (``OpBackupExport``)
will be extended to export instances to remote clusters.
#. Source cluster connects to destination cluster for each disk and
#. Source cluster removes the instance if requested
Instance move in pseudo code
-^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+++++++++++++++++++++++++++++
.. highlight:: python
.. highlight:: text
Miscellaneous notes
-^^^^^^^^^^^^^^^^^^^
++++++++++++++++++++
- A very similar system could also be used for instance exports within
the same cluster. Currently OpenSSH is being used, but could be
requirements.
+Privilege separation
+--------------------
+
+Current state and shortcomings
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+All Ganeti daemons are run under the user root. This is not ideal from a
+security perspective as for possible exploitation of any daemon the user
+has full access to the system.
+
+In order to overcome this situation we'll allow Ganeti to run its daemon
+under different users and a dedicated group. This also will allow some
+side effects, like letting the user run some ``gnt-*`` commands if one
+is in the same group.
+
+Implementation
+~~~~~~~~~~~~~~
+
+For Ganeti 2.2 the implementation will be focused on a the RAPI daemon
+only. This involves changes to ``daemons.py`` so it's possible to drop
+privileges on daemonize the process. Though, this will be a short term
+solution which will be replaced by a privilege drop already on daemon
+startup in Ganeti 2.3.
+
+It also needs changes in the master daemon to create the socket with new
+permissions/owners to allow RAPI access. There will be no other
+permission/owner changes in the file structure as the RAPI daemon is
+started with root permission. In that time it will read all needed files
+and then drop privileges before contacting the master daemon.
+
+
Feature changes
----------------
+===============
KVM Security
-~~~~~~~~~~~~
+------------
Current state and shortcomings
-++++++++++++++++++++++++++++++
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Currently all kvm processes run as root. Taking ownership of the
hypervisor process, from inside a virtual machine, would mean a full
option of subverting other basic services on the cluster (eg: ssh).
Proposed changes
-++++++++++++++++
+~~~~~~~~~~~~~~~~
We would like to decrease the surface of attack available if an
hypervisor is compromised. We can do so adding different features to
subvert the node.
Dropping privileges in kvm to a single user (easy)
-^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+++++++++++++++++++++++++++++++++++++++++++++++++++
By passing the ``-runas`` option to kvm, we can make it drop privileges.
The user can be chosen by an hypervisor parameter, so that each instance
- read unprotected data on the node filesystem
Running kvm in a chroot (slightly harder)
-^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
++++++++++++++++++++++++++++++++++++++++++
By passing the ``-chroot`` option to kvm, we can restrict the kvm
process in its own (possibly empty) root directory. We need to set this
Running kvm with a pool of users (slightly harder)
-^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+++++++++++++++++++++++++++++++++++++++++++++++++++
If rather than passing a single user as an hypervisor parameter, we have
a pool of useable ones, we can dynamically choose a free one to use and
can still be combined with the chroot benefits.
Running iptables rules to limit network interaction (easy)
-^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
These don't need to be handled by Ganeti, but we can ship examples. If
the users used to run VMs would be blocked from sending some or all
Running kvm inside a container (even harder)
-^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+++++++++++++++++++++++++++++++++++++++++++++
Recent linux kernels support different process namespaces through
control groups. PIDs, users, filesystems and even network interfaces can
just rely on iptables.
Implementation plan
-+++++++++++++++++++
+~~~~~~~~~~~~~~~~~~~
We will first implement dropping privileges for kvm processes as a
single user, and most probably backport it to 2.1. Then we'll ship
Finally we'll look into namespaces and containers, although that might
slip after the 2.2 release.
+New OS states
+-------------
+
+Separate from the OS external changes, described below, we'll add some
+internal changes to the OS.
+
+Current state and shortcomings
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+There are two issues related to the handling of the OSes.
+
+First, it's impossible to disable an OS for new instances, since that
+will also break reinstallations and renames of existing instances. To
+phase out an OS definition, without actually having to modify the OS
+scripts, it would be ideal to be able to restrict new installations but
+keep the rest of the functionality available.
+
+Second, ``gnt-instance reinstall --select-os`` shows all the OSes
+available on the clusters. Some OSes might exist only for debugging and
+diagnose, and not for end-user availability. For this, it would be
+useful to "hide" a set of OSes, but keep it otherwise functional.
+
+Proposed changes
+~~~~~~~~~~~~~~~~
+
+Two new cluster-level attributes will be added, holding the list of OSes
+hidden from the user and respectively the list of OSes which are
+blacklisted from new installations.
+
+These lists will be modifiable via ``gnt-os modify`` (implemented via
+``OpClusterSetParams``), such that even not-yet-existing OSes can be
+preseeded into a given state.
+
+For the hidden OSes, they are fully functional except that they are not
+returned in the default OS list (as computed via ``OpOsDiagnose``),
+unless the hidden state is requested.
+
+For the blacklisted OSes, they are also not shown (unless the
+blacklisted state is requested), and they are also prevented from
+installation via ``OpInstanceCreate`` (in create mode).
+
+Both these attributes are per-OS, not per-variant. Thus they apply to
+all of an OS' variants, and it's impossible to blacklist or hide just
+one variant. Further improvements might allow a given OS variant to be
+blacklisted, as opposed to whole OSes.
+
External interface changes
---------------------------
+==========================
+
+
+OS API
+------
+
+The OS variants implementation in Ganeti 2.1 didn't prove to be useful
+enough to alleviate the need to hack around the Ganeti API in order to
+provide flexible OS parameters.
+
+As such, for Ganeti 2.2 we will provide support for arbitrary OS
+parameters. However, since OSes are not registered in Ganeti, but
+instead discovered at runtime, the interface is not entirely
+straightforward.
+
+Furthermore, to support the system administrator in keeping OSes
+properly in sync across the nodes of a cluster, Ganeti will also verify
+(if existing) the consistence of a new ``os_version`` file.
+
+These changes to the OS API will bump the API version to 20.
+
+
+OS version
+~~~~~~~~~~
+
+A new ``os_version`` file will be supported by Ganeti. This file is not
+required, but if existing, its contents will be checked for consistency
+across nodes. The file should hold only one line of text (any extra data
+will be discarded), and its contents will be shown in the OS information
+and diagnose commands.
+
+It is recommended that OS authors increase the contents of this file for
+any changes; at a minimum, modifications that change the behaviour of
+import/export scripts must increase the version, since they break
+intra-cluster migration.
+
+Parameters
+~~~~~~~~~~
+
+The interface between Ganeti and the OS scripts will be based on
+environment variables, and as such the parameters and their values will
+need to be valid in this context.
+
+Names
++++++
+
+The parameter names will be declared in a new file, ``parameters.list``,
+together with a one-line documentation (whitespace-separated). Example::
+
+ $ cat parameters.list
+ ns1 Specifies the first name server to add to /etc/resolv.conf
+ extra_packages Specifies additional packages to install
+ rootfs_size Specifies the root filesystem size (the rest will be left unallocated)
+ track Specifies the distribution track, one of 'stable', 'testing' or 'unstable'
+
+As seen above, the documentation can be separate via multiple
+spaces/tabs from the names.
+
+The parameter names as read from the file will be used for the command
+line interface in lowercased form; as such, there shouldn't be any two
+parameters which differ in case only.
+
+Values
+++++++
+
+The values of the parameters are, from Ganeti's point of view,
+completely freeform. If a given parameter has, from the OS' point of
+view, a fixed set of valid values, these should be documented as such
+and verified by the OS, but Ganeti will not handle such parameters
+specially.
+
+An empty value must be handled identically as a missing parameter. In
+other words, the validation script should only test for non-empty
+values, and not for declared versus undeclared parameters.
+
+Furthermore, each parameter should have an (internal to the OS) default
+value, that will be used if not passed from Ganeti. More precisely, it
+should be possible for any parameter to specify a value that will have
+the same effect as not passing the parameter, and no in no case should
+the absence of a parameter be treated as an exceptional case (outside
+the value space).
+
+
+Environment variables
+^^^^^^^^^^^^^^^^^^^^^
+
+The parameters will be exposed in the environment upper-case and
+prefixed with the string ``OSP_``. For example, a parameter declared in
+the 'parameters' file as ``ns1`` will appear in the environment as the
+variable ``OSP_NS1``.
+
+Validation
+++++++++++
+
+For the purpose of parameter name/value validation, the OS scripts
+*must* provide an additional script, named ``verify``. This script will
+be called with the argument ``parameters``, and all the parameters will
+be passed in via environment variables, as described above.
+
+The script should signify result/failure based on its exit code, and
+show explanatory messages either on its standard output or standard
+error. These messages will be passed on to the master, and stored as in
+the OpCode result/error message.
+
+The parameters must be constructed to be independent of the instance
+specifications. In general, the validation script will only be called
+with the parameter variables set, but not with the normal per-instance
+variables, in order for Ganeti to be able to validate default parameters
+too, when they change. Validation will only be performed on one cluster
+node, and it will be up to the ganeti administrator to keep the OS
+scripts in sync between all nodes.
+
+Instance operations
++++++++++++++++++++
+
+The parameters will be passed, as described above, to all the other
+instance operations (creation, import, export). Ideally, these scripts
+will not abort with parameter validation errors, if the ``verify``
+script has verified them correctly.
+
+Note: when changing an instance's OS type, any OS parameters defined at
+instance level will be kept as-is. If the parameters differ between the
+new and the old OS, the user should manually remove/update them as
+needed.
+
+Declaration and modification
+++++++++++++++++++++++++++++
+
+Since the OSes are not registered in Ganeti, we will only make a 'weak'
+link between the parameters as declared in Ganeti and the actual OSes
+existing on the cluster.
+
+It will be possible to declare parameters either globally, per cluster
+(where they are indexed per OS/variant), or individually, per
+instance. The declaration of parameters will not be tied to current
+existing OSes. When specifying a parameter, if the OS exists, it will be
+validated; if not, then it will simply be stored as-is.
+
+A special note is that it will not be possible to 'unset' at instance
+level a parameter that is declared globally. Instead, at instance level
+the parameter should be given an explicit value, or the default value as
+explained above.
+
+CLI interface
++++++++++++++
+
+The modification of global (default) parameters will be done via the
+``gnt-os`` command, and the per-instance parameters via the
+``gnt-instance`` command. Both these commands will take an addition
+``--os-parameters`` or ``-O`` flag that specifies the parameters in the
+familiar comma-separated, key=value format. For removing a parameter, a
+``-key`` syntax will be used, e.g.::
+
+ # initial modification
+ $ gnt-instance modify -O use_dchp=true instance1
+ # later revert (to the cluster default, or the OS default if not
+ # defined at cluster level)
+ $ gnt-instance modify -O -use_dhcp instance1
+
+Internal storage
+++++++++++++++++
+
+Internally, the OS parameters will be stored in a new ``osparams``
+attribute. The global parameters will be stored on the cluster object,
+and the value of this attribute will be a dictionary indexed by OS name
+(this also accepts an OS+variant name, which will override a simple OS
+name, see below), and for values the key/name dictionary. For the
+instances, the value will be directly the key/name dictionary.
+
+Overriding rules
+++++++++++++++++
+
+Any instance-specific parameters will override any variant-specific
+parameters, which in turn will override any global parameters. The
+global parameters, in turn, override the built-in defaults (of the OS
+scripts).
+
.. vim: set textwidth=72 :
+.. Local Variables:
+.. mode: rst
+.. fill-column: 72
+.. End:
projects / ganeti-local / blobdiff