Reviewed-by: iustinp

Reviewed-by: iustinp

Reviewed-by: imsnah

This fixes removals of the pid files and clean exit in the init script

Reviewed-by: iustinp

There is no point in initializing sigchld_handler as an empty dict, when
it will be overwritten by an instance of utils.SignalHandler
immediately after.

Reviewed-by: iustinp

This patch enables the cluster modify to change:
  - enabled hypervisor list
  - hvparams (per hypervisor)
  - beparams (only the default group)

Syntax:
  gnt-cluster modify -B vcpus=3 -H xen-pvm:no_initrd_path

Validation for parameters is somewhat missing - the individual
hypervisors will be checked for syntax and validation, but beparams
doesn't have validation yes (nowhere), it should be added here once we
have a global method (will come soon).

Reviewed-by: imsnah

Requests are no longer logged to a separate file.

Reviewed-by: amishchenko

This is just a raw update without any special formatting.

Reviewed-by: imsnah

Reviewed-by: iustinp

Currently, if loading a job fails, the job queue code raises an
exception and prevents the proper processing of the jobs in the queue.
We change this so that unparseable jobs are instead archived (if not
already).

Reviewed-by: imsnah

If we don't actually know the current master (as determined via voting),
we prevent the failover.

The patch also changes some messages (capitalization, typos).

Reviewed-by: ultrotter

In order to account for future improvements to master failover, we move
the actual data gathering capabilities from ganeti-masterd into
bootstrap.py, and we leave only the verification into masterd.

The verification procedure is then changed to retry multiple times (up
to one minute) in case most nodes do not respond, and also the algorithm
is changed to require at least half (but not half+1) votes, since our
vote also should count (and we vote for ourselves).

Example for consistent (config-wise) cluster:
  - 5 node cluster, 2 nodes down: still start
  - 4 node cluster, 2 nodes down: retry for one minute, abort

Reviewed-by: ultrotter

This adds the set/reset in the jqueue and luxi modules, and a way to
query it in OpQueryConfigValues, and also the comand line interface for
it:
$ gnt-cluster queue info
The drain flag is unset
$ gnt-cluster queue drain
$ gnt-cluster queue info
The drain flag is set
$ gnt-cluster queue undrain
$ gnt-cluster queue info
The drain flag is unset

The choice of making the setting via luxi and not an opcode is that
opcodes can't be executed when drained, but we don't query via luxi
since in the future it might become a cluster property as opposed to a
node one.

Reviewed-by: imsnah

A new multi-node call is added that sets/resets the drain flag.

Reviewed-by: imsnah

We add a (per-node) queue drain flag that blocks new job submission.
There is not yet an interface to add/remove the flag (will come in next
patches).

Reviewed-by: imsnah

This patch adds a generic method to identify the ganeti error given its
class name, and implements this across the luxi protocol.

Reviewed-by: imsnah

Currently this is not derived from GenericError, but there's no reason
for it to be so.

Reviewed-by: imsnah

Reviewed-by: ultrotter

Reviewed-by: ultrotter

The patch adds a new ‘--no-wait-for-sync’ parameter to grow-disk similar
to the one in instance add, and changes the default to wait.

This is cleaner as at the moment when the command returns, we either
have a fully synced disk or there is an error.

This is a forward-port of rev 1183 on the 1.2 branch.

Reviewed-by: ultrotter

I'm not sure which parameters it should use, I took all I saw in the
code. Update: ultrotter says the list is OK for now.

Reviewed-by: ultrotter

Change the constant name to match the value (autobalance ->
auto_balance).

Also add the auto_balance header so that gnt-instance can list it.

Reviewed-by: ultrotter

The secondary nodes should not be checked for free memory in case of non
auto-balanced instances.

This is a forward-port from rev 1567 on the 1.2 branch.

Reviewed-by: ultrotter

This patch changes the cluster verify in two ways:
  - do not use instances marked with auto_balance=False in N+1 memory
    checks
  - report, if any, the number of instances with auto_balance=False

Forward port from rev 1562 on the 1.2 branch.

Reviewed-by: ultrotter

After the locking changes, and the partial locking of nodes, the
instance/node list is not ordered anymore. Reorder so that it behaves as
before.

Reviewed-by: ultrotter

Reviewed-by: ultrotter

This big patch changes the master code to use the beparams. Errors might
have crept in, but it passes a small burnin.

Reviewed-by: ultrotter

The backend.FinalizeExport function is changed to use the beparams
instead of the instance attributes. Future enhancements should be done
in order to export and import/reuse the whole be/hv params.

Reviewed-by: ultrotter

This patch changes the hypervisor functions to use the beparams dict
instead of the instance attributes.

Reviewed-by: ultrotter

As for the hvparams, we need to replace the sent instance beparams.

Reviewed-by: ultrotter

This is used to validate the parameters in both “gnt-instance add” and
“gnt-backup import”.

Reviewed-by: ultrotter

Memory is renamed for easier use.

Reviewed-by: ultrotter

When creating an instance, we need to check not the opcode hvparams, but
the final, filled hvparams against validity. While we do this for the
remote node calls (i.e. ValidateParams), we didn't do this for
CheckParameterSyntax.

Reviewed-by: imsnah

This patch adds a new '-s' parameter to ‘gnt-instance info’ that makes
it return only 'static' information. This is much faster, especially for
drbd instances.

This is a forward-port of rev 1570 on the ganeti-1.2 branch, resending
due to some conflicts.

Reviewed-by: imsnah

Some informations are not printed nicely (e.g. “virtual CDROM: False”),
but this is the first step.

Reviewed-by: imsnah

Reviewed-by: imsnah

This is just a change of the various hvm_ and pvm parameters to the hv
model. Parameters are queried via hv/$name or via the whole dict as
returned by hvparams.

Reviewed-by: ultrotter,imsnah

This big patch changes instance create to the new hvparams structure.
Old parameters are removed, so old jobs or old instances file will break
current clusters.

Reviewed-by: ultrotter

We have a problem with the current model of combining instance lists
from multiple hypervisors: we don't allow duplicates, but "xm list"
gives the same output for both pvm and hvm. This is a lack in the actual
xen hypervisor implementation/split between pvm and hvm, but for now we
implement a weak workaround: identical instance params will be allowed,
and merged. This breaks because there is a delta in listing, and should
be treated as temporary workaround only.

Note that there are two cases for duplicate instance: the above one (xen
is the same, whether pvm or hvm), and the other case, the real error,
when we have two different hypervisors reporting the same instance name.
The latter case needs to be handled better (not by refusing to list the
instances in the backend).

Reviewed-by: ultrotter

We fill the instance hvparams with cluster defaults, and send a modified
dict over the wire to the node methods - they don't know anything about
cluster defaults versus instance parameters.

Reviewed-by: ultrotter