Reviewed-by: iustinp

Reviewed-by: iustinp

Reviewed-by: iustinp

This patch does the following:
  - add constants.GANETI_RUNAS = "root", which is used to compute
    the homedir (and thus the .ssh directory) instead of hardcoding
    "/root/.ssh" in backend.AddNode and backend.LeaveCluster
  - add constants.SSH_CONFIG_DIR (currently hardcoded to /etc/ssh) that
    is used in backend instead of hardcoding it (preparation for
    selecting that at ./configure time)
  - some more internal cleanup in backend.AddNode

Reviewed-by: imsnah

Reviewed-by: schreiberal

Currently there are a few version of "write a file safely" in the code.
This patch adds a generic function that should be able to replace all
the other versions.

The "take attributes from the target file" option, while useful, is not
implemented by this patch.

Reviewed-by: imsnah,ultrotter

Since we remove only files from DATA_DIR and not from subdirectories,
let's not walk the entire tree, a simple listdir suffices. Also switch
to utils.RemoveFile from simple os.unlink.

Reviewed-by: imsnah

Currently GetHomeDir accepts UIDs only. Enhance it to accept either a
user name or a user id, to allow for nicer usage.

Reviewed-by: imsnah

Put some order in gnt-os putting helper functions above and command functions
below. Also add a new _DiagnoseOSName() function and use it to make _GetAllOS
simplier to understand. Change its name to _DiagnoseByOS which makes more clear
what it does.



Reviewed-by: iustinp

To do this we abstract the creation of the all_os dictionary, which was
produced in DiagnoseOS and use it both for listing and diagnosing.


Reviewed-by: iustinp

For the configuration update hook, it's useful to have a consistent name
for the target of the operation. As such, the LU code is modified to
include an GANETI_OP_TARGET that points either to the cluster (name),
node name or instance name depending on the opcode.

Also, the NoHooksLU is modified such that its build env method returns
an empty (but conformant) result. This should improve things in case by
mistake this class' BuildHooksEnv is called.

Reviewed-by: imsnah

This patch adds a special hook: the post-configuration update hook. This
hook has only a post phase that runs after a top-level LU that modified
the configuration.

Since the hook is a post-phase one, no error checking is done on the
results. The hook runs only on the master.

Reviewed-by: imsnah

This patch moves some of the environment processing from _BuildEnv to a
new _RunWrapper command which does the stringification and adds the
sstore variables.

The reasoning is that the sstore can be fresher than before the
execution (e.g.  in case of cluster init).

In order to support thise, we also need to modify cmdlib.LUInitCluster:
  - memorize the sstore and cfgw newly created in the Exec function
  - no need to build the custom environment in the BuildHooks

Currently, the HooksMaster creation and execution decision is in the
Processor class. This is not optimal, so we change to always create a
hooks master and instead make the decision inside that class, by
creating empty node lists for both pre and post if the lu doesn't
support hooks. This way, hooks decisions are moved to HooksMaster (where
they belong).

Reviewed-by: imsnah

The HooksMaster class doesn't use the cfg parameter, and it's better to
use it from the LU anyway (if needed). Let's remove it.

Also, the sstore of the LU can be fresher than the sstore we got at init
time, so use that instead and remove our own.

Reviewed-by: imsnah

This allows external callers to determine if the configuration has been
changed.

Reviewed-by: imsnah

Reviewed-by: iustinp

The beginning of the infrastructure for detailed instance restarts. Does
not affect behaviour if code yet.

Reviewed-by: iustinp

This removes the dependency on either the uuid module or e2fsprogs' uuidgen.

Reviewed-by: iustinp

This can be used to replace hardcoded "/root/" paths.

Reviewed-by: imsnah

This patch completely  gets rid of fping
 - replace all fping invocations with TcpPing calls
 - update documentation accordingly.
 - associated cleanups (use constant for localhost IP, use more sensible
   defaults for TcpPing and _use_ those)

Reviewed-by: iustinp

- Implement colours in qa_utils.
- Print warning for cron script.

Reviewed-by: iustinp

This adds:
  - fix the case when a dumb caller didn't stringify its values
  - explicitly raise a ProgrammerError in case a field is missing from
    the headers dictionary

Reviewed-by: imsnah

Currently, AskUser dies with -ESPIPE if the user gives more than one
character plus newline. This is because the python library, while
returning only two chars from the readline(2) call, will cache the rest
of the input, and when we do a write, it will try to seek back to just
after the last returned char. This fails on /dev/tty, so an exception is
raised. However, then opening the file descriptor in O_APPEND mode,
python will not issue such seeks, thereby fixing this problem.

The other alternative, opening in unbuffered mode, is not as good, since
then python will issue one-byte reads at a time, but the tty will read
the entire line, so at the next readline call, whatever remained in the
descriptor buffer is returned from the kernel to python, and the user
doesn't get a chance to enter something at all.

Reviewed-by: imsnah

Reviewed-by: iustinp

This patch adds a new 'evacuate' subcommand to gnt-node. The command
will do a replace disks for all instances having that node as secondary
with the new target being the new node given.

The syntax is:
  gnt-node evacuate src_node target_node

The command by itself doesn't do any resource checks, and instead relies
on the LUFailoverInstance code to do that.

Reviewed-by: imsnah

Reviewed-by: iustinp

Reviewed-by: iustinp

Since modules are not directly executables, remove the shebang from
them. This helps with lintian warnings.

Also make the autogenerated _autoconf.py contain two comment lines at
the beginning, like the other modules.

Reviewed-by: ultrotter

- Change format of watcher state file to JSON.
- Move log path for watcher script to constants.py.

Reviewed-by: iustinp

This patch implements a simple failover command for failing over all
primary instances. This is just a batched form of gnt-instance failover.

Reviewed-by: imsnah

Reviewed-by: ultrotter

The new version debugs Hidden OSes as well.


Reviewed-by: iustinp

This was introduced in rev 208.

Reviewed-by: imsnah

This patch moves the error formatting into a separate function that will
be exported so that scripts that execute multiple opcodes can create the
same output as for GenericMain. Also, GenericMain becomes more readable.

Reviewed-by: imsnah

As far as I understand, this was just a debugging aid. Remove it so that
configuration errors are nicely handled.

Reviewed-by: imsnah

This patch allows the SubmitOpcode function to take a pre-created
Processor instance and a different feedback function. This is helpful
when submitting more than one opcode from a shell script.

Reviewed-by: imsnah

Reviewed-by: iustinp

This patch adds a simple handler for TagError that prints the error
(better than having a stack dump).

Reviewed-by: imsnah

This patch adds a '--from' option to the {add,remove}-tags commands that
allows the command to read from a file or from stdin (when passing a
dash for the filename).

Reviewed-by: imsnah