On cluster init if the user specifies a default hypervisor (with -t)
which is not in the default list of enabled hypervisors (currently just
xen-pvm) without explicitely specifying the list we silently override
the choice.

With this patch we set the list by default to just the required one, and
we bail out should the list be hand-specified and not contain the
default one. This still has an issue when the user doesn't specify a
default hypervisor but specifies a list which doesn't include xen-pvm:
in this case though we give an error, rather than silently discarding
choices.

Reviewed-by: imsnah

Where we used/forgot to validate beparams we now use the new common function.

Reviewed-by: imsnah

This function will be used in LUCreateInstance, LUSetInstanceParams,
LUSetClusterParams and InitCluster to check the backend parameters
validity and convert the relevant values to integer, without duplicating
code. It lives in utils as bootstrap.py is calling it too.

Reviewed-by: imsnah

Reviewed-by: imsnah

When a value is set to constants.VALUE_DEFAULT we have to remove it from
the specific instance dict, as this way it will be populated from the
cluster before. If instead it's specified as constants.VALUE_NONE we'll
explicitely set it to None, to override its presence with a different
values in such defaults. However, currently, we handle None values only
for hvparams, that have a real use case for them.

Reviewed-by: imsnah

If the hv/be parameter lowercase value is set to "default" we'll pass
constants.VALUE_DEFAULT, if it's set to "none" we'll pass
constants.VALUE_NONE.

Reviewed-by: imsnah

--src-node and --src-dir are not mandatory anymore

Reviewed-by: iustinp

If src_node is not there we'll default to using the currently exported
instance name as src_path. Also, if src_path is not absolute we'll look
for it in EXPORT_DIR.

Reviewed-by: iustinp

If we get called with no source node we'll thread src_path as an
instance name exported in EXPORT_DIR in one of the nodes and look for
it with the export_list rpc call.

Reviewed-by: iustinp

Currently the node lock also guards against removing the import at the
wrong time, so if we're importing an instance image we want to keep the
source node locked. In the future we might want to put export locks at a
different level than node locks.

Reviewed-by: iustinp

The ssconf files were not updated by the master failover. We need to
push them, and since we already have RPC initialized, we can use the
standard ConfigWriter to do so - this will take care of both the config
file and the ssconf files.

Reviewed-by: imsnah

Currently cluster verify checks all nodes for the same set of files,
even if the nodes are not master candidates.

This patch adds back checking of ssconf files for consistency and splits
the checksum check into different error reporting messages based on
candidate role.

Reviewed-by: imsnah

Reviewed-by: imsnah

In gnt-cluster modify we prevent demotion from the candidate role if
there are not enough master candidates left.

Reviewed-by: imsnah

This patch adds a new cluster paramater "candidate_pool_size" which
tracks the desired size of the list of nodes with the master_candidate
flag set.

Reviewed-by: imsnah

Reviewed-by: imsnah

Reviewed-by: imsnah

This patch restricts the job propagation to master candidates only, by
not registering non-candidates in the job queue node lists.

Note that we do intentionally purge the job queue if a node is toggled
to non-master status.

Reviewed-by: imsnah

This patch restricts the config data replication to master candidates
only.

Reviewed-by: imsnah

This patch adds the OpCode, LogicalUnit and gnt-node command for
modifying node parameters, more specifically the master candidate flag
for a node.

Reviewed-by: imsnah

This patch adds listing of the master_candidate field (as Y/N) and of
the master role (again Y/N) for nodes.

Reviewed-by: imsnah

The field is not yet used.

Reviewed-by: imsnah

We have (again) the KeyToFilename function, so we move the writing of
the files to a method under SimpleStore.

Reviewed-by: imsnah

This patch adds node_list in the list of replicated values from
ConfigWriter.

Reviewed-by: imsnah

This partially reverts the "Get rid of ssconf" patch.

It adds back a simpler version of the SimpleStore class, and drops the
WritableSimpleStore class. The new version of the class also has
node_list as a new key, and increases the size of the keys so that big
clusters will fit the node list. Also, the SS_* constants are moved to
constants.py, since the ConfigWriter class will need them too in order
to generate the values dictionary.

It also changes the GetMasterAndMyself function to use the SimpleStore
by default, and the backend._GetConfig to use it too (it has all the
needed keys).

Reviewed-by: imsnah

This allows burnin to work with diskless instances (since we cannot pass
right now no disks to it).

Reviewed-by: imsnah

There are still issues, especially with "gnt-instance modify" and
resetting values. However, this is a start.

Reviewed-by: ultrotter

They're not used anymore.

Reviewed-by: ultrotter

Unfortunately, a rpc.Client object was passed as the first parameter,
causing the function to always fail.

Found during QA testing.

Reviewed-by: ultrotter

Currently it was left alone, and thus its value was "null".

Reviewed-by: imsnah

[Forward-port from the 1.2 branch]

If ganeti starts before xend, the node information will not have all the
fields filled in. The patch changes so that missing keys will be treated
as unknown (this applies to other cases as well, not only xend not
started).

Reviewed-by: ultrotter

Reviewed-by: imsnah

Reviewed-by: imsnah

Reviewed-by: iustinp

Reviewed-by: iustinp

- Add new "canceling" status
- Notify clients when job is canceled
- Give a return value from CancelJob
- Handle it in the client library

Reviewed-by: iustinp

Currently, the node add operation uses a job to query the node name and
the bootstrap function directly reads the config file for the cluster
name.

This patch changes to that both the cluster name and the verification of
the node is done via queries to the master.

Reviewed-by: ultrotter

In revision 2072 "ConfigWriter: change cluster serial meaning" I misread
the serial_no update logic: it was about updating the serial number on
the object itself, not on the cluster.

So we don't actually have at all cluster serial number increase when a
node is changed (not removed/added).

This patch revers to the original always increase the target serial
number and adds increase of the cluster serial number in case a node has
been changed.

Reviewed-by: ultrotter

Dictionaries raise KeyError, and not ValueError when invalid keys are
passes to del.

Reviewed-by: imsnah

Currently, we increase the cluster serial number for instance additions,
removals and renames. This is conforming with the REST paradigm, however
it means that for each of these operations, we need to push ssconf
updates to all nodes.

In order to support future cases with reduced set of master-eligible
nodes, we want to reduce the ssconf pushes (which need to be to all
nodes). This patch changes the meaning for the cluster serial number so
that it doesn't track instance operations at all.

This means that addition of an instance can fail due to concurrent
additions, even if the cluster serial has not changed. It slightly
breaks the REST paradigm, but IMHO it's better for actual usage.

Reviewed-by: ultrotter