Many burn steps to a manual check of instance aliveness, via duplicate
code. This patch moves this code to a decorator.

Signed-off-by: Iustin Pop <iustin@google.com>
Reviewed-by: Michael Hanselmann <hansmi@google.com>

Some burnin steps are idempotent: e.g. reinstalling an instance (from
burning p.o.v.) can be done multiple times without any side-effects that
would affect later burnin steps. As such, failing the whole burnin
process due a reinstall failure is undesirable.

This patch modifies burnin by marking each opcode (in case of individual
execution) and job set retryable or not. Retryable actions will be
retried up to a number of times, after which we give up and return
failure.

One side-effect is that in case of full-failure in retryable job sets we
lose the original exception (but we do log its string format), so we
have a little bit less information in this case.

Signed-off-by: Iustin Pop <iustin@google.com>
Reviewed-by: Guido Trotter <ultrotter@google.com>

Signed-off-by: Michael Hanselmann <hansmi@google.com>
Reviewed-by: Iustin Pop <iustin@google.com>

A long-standing bug in burnin makes errors during the removal phase
(e.g. because an import has failed, or because the initial creation has
failed) hide the original error.

This patch suppresses removal errors if we are already in ‘has_err’
mode, and otherwise it displays them normally.

Signed-off-by: Iustin Pop <iustin@google.com>
Reviewed-by: Guido Trotter <ultrotter@google.com>

The list of upload files is built currently at every UploadFile() call.
This patch moves it to a separate variable which is initialized only
once.

This won't make much difference but I regard it as cleanup.

Signed-off-by: Iustin Pop <iustin@google.com>
Reviewed-by: Guido Trotter <ultrotter@google.com>

Commit 55efe6da "Convert instance
reinstall to multi instance model" actually broke instance reinstall for
single-instance cases. This one-liner fixes it.

Signed-off-by: Iustin Pop <iustin@google.com>
Reviewed-by: Guido Trotter <ultrotter@google.com>
(cherry picked from commit b6e243ab)

It seems epydoc needs fully-qualified references, and doesn't deal with
relative ones (not even in the current module) if there are any
ambiguities.

There are other epydoc warnings, in the rapi docstrings, but those are
left as-is as they're removed in 2.1.

Signed-off-by: Iustin Pop <iustin@google.com>
Reviewed-by: Guido Trotter <ultrotter@google.com>

Currently, unclean master daemon shutdown overwrites all of a job's
opcode status and result with error/None. This is incorrect, since the
any already finished opcode(s) should have their status and result
preserved, and only not-yet-processed opcodes should be marked as
‘error’. Cancelling jobs between opcodes does the same (but this is not
allowed currently by the code, so it's not as important as unclean
shutdown).

This patch adds a new _QueuedJob function that only overwrites the
status and result of finalized opcodes, which is then used in job queue
init and in the cancel job functions. The patch also adds some comments
and a new set constants in constants.py highlighting the finalized vs.
non-finalized opcode statuses.

Signed-off-by: Iustin Pop <iustin@google.com>
Reviewed-by: Guido Trotter <ultrotter@google.com>

Currently gnt-debug submits jobs individually, but in 2.1 JobExecutor
uses the optimized SubmitManyJobs luxi call and as such should be used
whenever multiple jobs need to be submitted.

This patch converts gnt-debug submit-job to use it and also removes an
extra empty line in the JobExecutor class.

Signed-off-by: Iustin Pop <iustin@google.com>
Reviewed-by: Guido Trotter <ultrotter@google.com>

This patch converts ‘gnt-instance reinstall’ from single-instance to
multi-instance model; since this is dangerours, it's required to pass
“--force --force-multiple” to skip the confirmation.

Signed-off-by: Iustin Pop <iustin@google.com>
Reviewed-by: Guido Trotter <ultrotter@google.com>
(cherry picked from commit 55efe6da)

This small patch changed the batch create functionality to use the job
executor instead of single-job submits.

Signed-off-by: Iustin Pop <iustin@google.com>
Reviewed-by: Guido Trotter <ultrotter@google.com>
(cherry picked from commit d4dd4b74)

This patch changes the generic "multiple job executor" to use the many
jobs submit model, which automatically makes all its users use the new
model.

This makes, for example, startup/shutdown of a full cluster much more
logical (all the submitted job IDs are visible fast, and then waiting
for them proceeds normally).

Signed-off-by: Iustin Pop <iustin@google.com>
Reviewed-by: Guido Trotter <ultrotter@google.com>
(cherry picked from commit 23b4b983)

As a workaround for the job submit timeouts that we have, this patch
adds a new luxi call for multi-job submit; the advantage is that all the
jobs are added in the queue and only after the workers can start
processing them.

This is definitely faster than per-job submit, where the submission of
new jobs competes with the workers processing jobs.

On a pure no-op OpDelay opcode (not on master, not on nodes), we have:
  - 100 jobs:
    - individual: submit time ~21s, processing time ~21s
    - multiple:   submit time 7-9s, processing time ~22s
  - 250 jobs:
    - individual: submit time ~56s, processing time ~57s
                  run 2:      ~54s                  ~55s
    - multiple:   submit time ~20s, processing time ~51s
                  run 2:      ~17s                  ~52s

which shows that we indeed gain on the client side, and maybe even on
the total processing time for a high number of jobs. For just 10 or so I
expect the difference to be just noise.

This will probably require increasing the timeout a little when
submitting too many jobs - 250 jobs at ~20 seconds is close to the
current rw timeout of 60s.

Signed-off-by: Iustin Pop <iustin@google.com>
Reviewed-by: Guido Trotter <ultrotter@google.com>
(cherry picked from commit 2971c913)

If a job with more than one opcodes is being processed, and the master
daemon crashes between two opcodes, we have the first N opcodes marked
successful, and the rest marked as queued. This means that the overall
jbo status is queued, and thus on master daemon restart it will be
resent for completion.

However, the RunTask() function in jqueue.py doesn't deal with
partially-completed jobs. This patch makes it simply skip such opcodes.

An alternative option would be to not mark partially-completed jobs as
QUEUED but instead RUNNING, which would result in aborting of the job at
restart time.

Signed-off-by: Iustin Pop <iustin@google.com>
Reviewed-by: Guido Trotter <ultrotter@google.com>

In case the job fails, we try to set the job's run_op_idx to -1.
However, this is a wrong variable, which wasn't detected until the
__slots__ addition. The correct variable is run_op_index.

Signed-off-by: Iustin Pop <iustin@google.com>
Reviewed-by: Guido Trotter <ultrotter@google.com>

Adding slots to _QueuedOpCode decreases memory usage (of these objects)
by roughly four times. It is a lesser change for _QueuedJobs.

Signed-off-by: Iustin Pop <iustin@google.com>
Reviewed-by: Michael Hanselmann <hansmi@google.com>

Signed-off-by: Michael Hanselmann <hansmi@google.com>
Reviewed-by: Iustin Pop <iustin@google.com>

This patch converts the opcode loading to a pre-built map (at import
time) instead of iteration over the globals dict at each call.

Microbenchmarks show that this should be around three times faster, and
burnin still passes.

Signed-off-by: Iustin Pop <iustin@google.com>
Reviewed-by: Michael Hanselmann <hansmi@google.com>

Signed-off-by: Iustin Pop <iustin@google.com>
Reviewed-by: Olivier Tharan <olive@google.com>

* master:
  Update NEWS and version for 2.0.2 release
  Improve the description of node flags in man page
  Change default stripe count to 1
  Use full-stripe size in LVM growth
  RAPI: implement instance reinstall

Signed-off-by: Iustin Pop <iustin@google.com>
Reviewed-by: Michael Hanselmann <hansmi@google.com>

Signed-off-by: Iustin Pop <iustin@google.com>
Reviewed-by: Guido Trotter <ultrotter@google.com>

[iustin@google.com: slightly reworded the explanation for offline and
changed the commit message]
Signed-off-by: Iustin Pop <iustin@google.com>
Reviewed-by: Iustin Pop <iustin@google.com>

This parameter is now mandatory for the cluster config to work.

Signed-off-by: Guido Trotter <ultrotter@google.com>
Reviewed-by: Michael Hanselmann <hansmi@google.com>

- Check that the enabled hypervisors list is valid
- Check that the master node is a valid node

Signed-off-by: Guido Trotter <ultrotter@google.com>
Reviewed-by: Iustin Pop <iustin@google.com>

Signed-off-by: Guido Trotter <ultrotter@google.com>
Reviewed-by: Iustin Pop <iustin@google.com>

In order not to change the default during a stable series, we modify
configure.ac to default to one stripe, in effect keeping the status quo
(well, minus the LVM Attach() changes).

Signed-off-by: Iustin Pop <iustin@google.com>
Reviewed-by: Guido Trotter <ultrotter@google.com>

LVM has issues when growing stripped volumes, so it's best to specify
the growth in exact multiples of the full stripe size (as precise as
possible). For this we need to do a couple of changes:
  - in LVM Attach(), we query additionally the VG extent size and the LV
    stripe count; since this makes lvs return a (possibly) multi-line
    output, we now split it into lines and only take the last one
  - in LVM Grow(), we round up the increase in multiples of the full
    stripe size

The patch also sets the correct target size in DRBD growth.

Signed-off-by: Iustin Pop <iustin@google.com>
Reviewed-by: Olivier Tharan <olive@google.com>

It's been replaced by a simpler bootstrap.InitConfig function, which
does the same job, and is currently unused.

Signed-off-by: Guido Trotter <ultrotter@google.com>
Reviewed-by: Iustin Pop <iustin@google.com>

This function is not used.

Signed-off-by: Guido Trotter <ultrotter@google.com>
Reviewed-by: Michael Hanselmann <hansmi@google.com>

Currently if a disk is added later the base_index is not considered, and
all the disks are called disk0. This patch fixes it.

Signed-off-by: Guido Trotter <ultrotter@google.com>
Reviewed-by: Michael Hanselmann <hansmi@google.com>

SimpleStore is a lot less heavyweight than SimpleConfigReader, and to
just get the master name we can use that. This is the only usage of
SimpleConfigReader currently, but we're not going to delete the class,
as new usages will come in for ganeti-confd (in 2.1). Using it there,
though, will make the class even more heavy to load, so it makes sense
for this simple usage to be converted.

Signed-off-by: Guido Trotter <ultrotter@google.com>

This was broken by my pylint fixes patch.

Signed-off-by: Michael Hanselmann <hansmi@google.com>
Reviewed-by: Iustin Pop <iustin@google.com>

This patch adds instance reinstall to RAPI, with two optional parameters:
  - ‘os', in order to change the OS on reinstall
  - ‘nostartup’, in order to leave the instance down after reinstall

The call will first shutdown the instance, the reinstall it, and unless
‘nostartup’ has been passed and is equal to 1, it will be started
automatically.

Signed-off-by: Iustin Pop <iustin@google.com>
Reviewed-by: Guido Trotter <ultrotter@google.com>

* master:
  Create a new --no-voting option for masterfailover
  ganeti-masterd: allow non-interactive --no-voting

This allows failing over in certain corner cases, such as a 2 node
cluster with one node down. The man page is also updated to document
this dangerous option and how to recover from this situation.

Signed-off-by: Guido Trotter <ultrotter@google.com>
Reviewed-by: Iustin Pop <iustin@google.com>

This will be used by ganeti-noded to start ganeti-masterd in a
--no-voting masterfailover.

Signed-off-by: Guido Trotter <ultrotter@google.com>
Reviewed-by: Iustin Pop <iustin@google.com>

Signed-off-by: Michael Hanselmann <hansmi@google.com>
Reviewed-by: Iustin Pop <iustin@google.com>

Signed-off-by: Michael Hanselmann <hansmi@google.com>
Reviewed-by: Iustin Pop <iustin@google.com>

Signed-off-by: Michael Hanselmann <hansmi@google.com>
Reviewed-by: Iustin Pop <iustin@google.com>