valgrind: suppression for md_config_t string leak

Fixes: http://tracker.ceph.com/issues/17924
Signed-off-by: Casey Bodley <cbodley@redhat.com>

Merge pull request #990 from ceph/wip-suite-repo

Allow using an alternate suite repo

Reviewed-by: Dan Mick <dmick@redhat.com>

Merge pull request #994 from ceph/wip-shaman-default

Default use_shaman to True

Default use_shaman to True

Signed-off-by: Zack Cerza <zack@redhat.com>

valgrind.supp: xenial libstdc++ leak

Signed-off-by: Sage Weil <sage@redhat.com>

Merge pull request #993 from ceph/wip-rerun-0match

suite: Correct rerun behavior when no jobs match

suite: Correct rerun behavior when no jobs match

Also remove a stray print statement

Signed-off-by: Zack Cerza <zack@redhat.com>

Merge pull request #992 from ceph/wip-kernel-reboot-xenial

kernel: fix reboot into the new kernel on xenial

Merge pull request #987 from ceph/wip-misdirected-ops

ceph.conf: osd debug  misdirected ops = true

Merge pull request #969 from jcsp/wip-generate-caps

misc: remove generate_caps

worker: Use suite_relpath if set

Signed-off-by: Zack Cerza <zack@redhat.com>

run: Use suite_relpath if set

Signed-off-by: Zack Cerza <zack@redhat.com>

suite: add --suite-relpath option

Signed-off-by: Zack Cerza <zack@redhat.com>

worker: Use 'suite_repo' in job config if present

Signed-off-by: Zack Cerza <zack@redhat.com>

suite: add --suite-repo option

Signed-off-by: Zack Cerza <zack@redhat.com>

repo_utils: Use shallow clones of git repos

Signed-off-by: Zack Cerza <zack@redhat.com>

repo_utils: Give cloned repos more unique names

Signed-off-by: Zack Cerza <zack@redhat.com>

kernel: fix reboot into the new kernel on xenial

shutdown is a symlink to systemctl on xenial, and on "shutdown -r now"
network gets shut down before ssh has a chance to close gracefully.
There have been multiple complaints about this and some distantly
related fixes have made it into recent systemds [1].

This doesn't seem to happen on centos, probably because of different
systemd versions or sshd configuration, so this band-aid is for ubuntu
only.  "systemctl --no-block reboot" is nicer, but not available on
trusty.

[1] https://github.com/systemd/systemd/issues/2390

Signed-off-by: Ilya Dryomov <idryomov@gmail.com>

Merge pull request #991 from athanatos/wip-rerun

matrix.py: actually generate sorted descriptons

matrix.py: actually generate sorted descriptons

The existing code compares strings and tuples which, based on what seems
to be happening, is at best counterintuitive and at worst
non-deterministic.

Signed-off-by: Samuel Just <sjust@redhat.com>

Merge pull request #988 from ceph/wip-shaman-repo

ShamanProject: don't write broken repos

Reviewed-by: Dan Mick <dmick@redhat.com>

misc: avoid a circular import with .packaging

Import DEFAULT_OS_VERSION from .orchestra.opsys, where it is originally
defined.

Signed-off-by: Zack Cerza <zack@redhat.com>

ShamanProject: don't write broken repos

We saw several cases where repo file contents were actually small HTML
files containing "502 Bad Gateway" errors. This will cause ShamanProject
to raise an exception if that happens - and not write broken repo files.

Signed-off-by: Zack Cerza <zack@redhat.com>

Merge pull request #940 from ceph/wip-rh-changes

rh specific changes

ceph.conf: osd debug  misdirected ops = true

Signed-off-by: Sage Weil <sage@redhat.com>

Merge pull request #985 from ceph/wip-shaman-kernel

kernel task support for shaman

task.kernel: Support fetching kernels from Shaman

Signed-off-by: Zack Cerza <zack@redhat.com>

Add ShamanProject.scm_version

This new property returns .['extra]['version'] from the Shaman query
result.

Signed-off-by: Zack Cerza <zack@redhat.com>

Remove rh prefix from function names

remove rh-prefix from funciton names and
fix review comments

Signed-off-by: Vasu Kulkarni <vasu@redhat.com>

unit tests for install function

Signed-off-by: Vasu Kulkarni <vasu@redhat.com>

uninstall based on packages defined in downstream.yaml

Signed-off-by: Vasu Kulkarni <vasu@redhat.com>

function to install debian downstream packages

Install debian downstream packages based on packages
defined in rh/downstream.yaml defined in suites dir

Signed-off-by: Vasu Kulkarni <vasu@redhat.com>

function to setup debian downstream repo

Signed-off-by: Vasu Kulkarni <vasu@redhat.com>

use downstream.yaml to install packages and version check

remove hardcoded checks instead use downstream.yaml
defined in suites dir rh/downstream.yaml
for packages and version checks

Signed-off-by: Vasu Kulkarni <vasu@redhat.com>

use rh/downstream.yaml to define packages

Remove hardcoded checks, use rh/downstream.yaml
from suite dir to define packages to install

Signed-off-by: Vasu Kulkarni <vasu@redhat.com>

Merge pull request #984 from ceph/wip-dash-m

lock: -m alias for --machine-type

lock: -m alias for --machine-type

Just like suite.py.

Signed-off-by: Sage Weil <sage@redhat.com>

Merge pull request #983 from ceph/wip-local-kernel-regex

kernel: update get_sha1_from_pkg_name() regex

kernel: update get_sha1_from_pkg_name() regex

ceph-build kernel packages will be named -ceph-g${SHA1:0:12}.  As the
first step towards using ceph-build, change the local package regex.
With this, I'm able to download post ceph/ceph-build#540 packages and
install them as a local deb/rpm.

Signed-off-by: Ilya Dryomov <idryomov@gmail.com>

Merge pull request #980 from jcsp/wip-python-rgw

teuthology/nuke: update for python-rgw

Reviewed-by: Zack Cerza <zack@redhat.com>
Reviewed-by: Kefu Chai <kchai@redhat.com>

Merge pull request #981 from ceph/wip-unpin-ansible

Revert "Pin ansible to <2.2.0"

teuthology/nuke: update for python-rgw

Signed-off-by: John Spray <john.spray@redhat.com>

Revert "Pin ansible to <2.2.0"

This reverts commit f2576e7a719a129a3c8be603bd9993d2b1ec94f2.

We finally heeded all the deprecation warnings.

Signed-off-by: Zack Cerza <zack@redhat.com>

Merge pull request #979 from tchaikov/wip-remove-libcephfs

teuthology: update nuke for libcephfs2

Reviewed-by: Brad Hubbard <bhubbard@redhat.com>

teuthology: update nuke for libcephfs2

try to remove libcephfs2 as well.

Signed-off-by: Kefu Chai <kchai@redhat.com>

Merge pull request #974 from ceph/wip-install-subpackage

Make the install task a subpackage

Reviewed-by: Dan Mick <dmick@redhat.com>

ShamanProject: add repo installation and removal

Signed-off-by: Zack Cerza <zack@redhat.com>

Move repo installation to GitbuilderProject

Signed-off-by: Zack Cerza <zack@redhat.com>

ceph.conf: bdev debug aio = true

detect aio stalls

Signed-off-by: Sage Weil <sage@redhat.com>

Add NoRemoteError

Signed-off-by: Zack Cerza <zack@redhat.com>

Create a task.install.redhat submodule

This will hold most of the code specific to RHCS

Signed-off-by: Zack Cerza <zack@redhat.com>

Create a task.install.deb submodule

This will hold most of the code specific to Debian-based systems

Signed-off-by: Zack Cerza <zack@redhat.com>

Create a task.install.rpm submodule

This will hold most of the code specific to RPM-based systems

Signed-off-by: Zack Cerza <zack@redhat.com>

Create a task.install.util submodule

This will hold miscellaneous code that doesn't need to be top-level and
isn't distro-specific

Signed-off-by: Zack Cerza <zack@redhat.com>

Linter fixes

Signed-off-by: Zack Cerza <zack@redhat.com>

Make the install task a subpackage

Signed-off-by: Zack Cerza <zack@redhat.com>

Merge pull request #977 from ceph/wip-fix-vm

Move edit_sudoers.sh to task/internal/

Merge pull request #978 from SUSE/wip-fix-suse-user-data

openstack: fix location of network script in openSUSE user-data

openstack: fix location of network script in openSUSE user-data

Signed-off-by: Nathan Cutler <ncutler@suse.com>

Move edit_sudoers.sh to task/internal/

Signed-off-by: Zack Cerza <zack@redhat.com>

Merge pull request #975 from ceph/wip-install-excludes

install: apply exclude_packages globally

Reviewed-by: Dan Mick <dmick@redhat.com>

install: apply exclude_packages globally

Previously, we were only applying the setting for upgrades

Signed-off-by: Zack Cerza <zack@redhat.com>

nuke: Call the check_lock function, not the module

Signed-off-by: Zack Cerza <zack@redhat.com>

Merge pull request #823 from ceph/wip-internal

Task module refactoring

Reviewed-by: Dan Mick <dmick@redhat.com>

ceph.conf: mon allow pool delete = true

Signed-off-by: Sage Weil <sage@redhat.com>

Merge pull request #973 from dmick/wip-pull-directory

Change pull_directory to use a remote tar streaming to the local extract

Merge pull request #970 from ceph/wip-shaman-tags

ShamanProject: query tags

Reviewed-by: Dan Mick <dmick@redhat.com>

Use remote.get_tar_stream in misc.pull_directory

Avoids remote and local temp files, and subsequent read/write
thrashing of local dir.  Since some of those directories are
multiple tens or hundreds of gigabytes, the extra file load
is important.

Signed-off-by: Dan Mick <dan.mick@redhat.com>

Add orchestra.remote.get_tar_stream(); like get_tar but no tmpfile

Avoid creating a temporary file either on the remote or on the
local host; just open a connection to a remote tar and let 'er
rip.  Intended to be used by get_directory to avoid temp files
and local read/write hammering.

Signed-off-by: Dan Mick <dan.mick@redhat.com>

misc.pull_directory: fix misplaced log message, remove useless continue

Signed-off-by: Dan Mick <dan.mick@redhat.com>

Move task.internal.syslog to its own module

Signed-off-by: Zack Cerza <zack@redhat.com>

Move internal.vm_setup to its own module

Signed-off-by: Zack Cerza <zack@redhat.com>

Move internal.check_lock to its own module

Signed-off-by: Zack Cerza <zack@redhat.com>

Move internal.lock_machines to its own module

Signed-off-by: Zack Cerza <zack@redhat.com>

Make the internal task a subpackage

Signed-off-by: Zack Cerza <zack@redhat.com>

run_tasks: support tasks that are subpackages

Signed-off-by: Zack Cerza <zack@redhat.com>

Pin ansible to <2.2.0

Until we fix the deprecations

Signed-off-by: Zack Cerza <zack@redhat.com>

ShamanProject: translate tags to sha1s

Since Shaman doesn't want to have to know about tags, let's do the
translation from tag->sha1 ourselves using 'git ls-remote'.

Signed-off-by: Zack Cerza <zack@redhat.com>

Merge pull request #965 from ceph/wip-exiter

Don't leak console-logging processes when a job is killed

Reviewed-by: Dan Mick <dmick@redhat.com>

Merge pull request #971 from ceph/wip-ceph-mgr

teuthology: update nuke for ceph-mgr

teuthology: update nuke for ceph-mgr

Signed-off-by: John Spray <john.spray@redhat.com>

Move suite.util.build_git_url to repo_utils

Signed-off-by: Zack Cerza <zack@redhat.com>

Add repo_utils.ls_remote()

... by moving most of suite.util.git_ls_remote's code to that module.

Signed-off-by: Zack Cerza <zack@redhat.com>

Ensure console loggers die when the job does

Signed-off-by: Zack Cerza <zack@redhat.com>

Use psutil.Popen instead of subprocess.Popen

psutil.Popen is identical other than offering nice features like making
sure a PID hasn't been reused before it sends signals to it.

Signed-off-by: Zack Cerza <zack@redhat.com>

Revert "ceph.conf: debug monc = 1"

This reverts commit 5137d3c6cfd065c69fe7e3bb570ef0fc795febcb.

misc: remove generate_caps

This is moved to ceph-qa-suite.

Fixes: http://tracker.ceph.com/issues/17614
Signed-off-by: John Spray <john.spray@redhat.com>

Merge pull request #967 from ceph/wip-fedora-python

downburst: Make sure python 2 is installed on Fedora

Reviewed-by: Dan Mick <dmick@redhat.com>

Merge pull request #964 from ceph/wip-ssh-config

orchestra.connect: do not override user/host from ~/.ssh/config

Make sure python 2 is installed on Fedora

We need this for ansible.

Signed-off-by: Zack Cerza <zack@redhat.com>

Merge pull request #966 from ceph/wip-archive-default

Update the default archive_base path

Update the default archive_base path

None of our labs are using /var/lib/teuthworker/archive anymore. The
incorrect default could trip up users.

Signed-off-by: Zack Cerza <zack@redhat.com>

ceph.conf: debug monc = 1

Signed-off-by: Sage Weil <sage@redhat.com>

Add exit module

This module contains a helper to add functions to be run when a given
signal is received. These signal handlers ought to do as little as
possible to avoid interfering with the process shutting down.

Signed-off-by: Zack Cerza <zack@redhat.com>

orchestra.connect: do not override user/host from ~/.ssh/config

I don't know why this was ever the case, but ignoring user@host
from the targets list just seems like a bad idea, and causes
problems, because tasks/ansible/etc. assume that the user to use
is ubuntu.  If that's not what .ssh/config says, there's just
going to be trouble...but there's no reason to pay attention
to .ssh/config there; if you really want to play tricks, edit
your target list accordingly.

Also, the original code was apparently not prepared for the
key_filename to be a list, which it apparently always is,
at least now; handle that too.

Fixes: http://tracker.ceph.com/issues/17125
Signed-off-by: Dan Mick <dan.mick@redhat.com>

Merge pull request #963 from ceph/wip-rerun

Add rerun capability to teuthology-suite

Reviewed-by: Dan Mick <dmick@redhat.com>

suite: Add rerun capability

Using a new --rerun argument for teuthology-suite, optionally paired
with --rerun-statuses, we can now ask teuthology to re-schedule a
suite-run. The main use-case for the feature is to re-schedule only
those jobs from a given run which failed.

Fixes: #10945
http://tracker.ceph.com/issues/10945

Signed-off-by: Zack Cerza <zack@redhat.com>

Add ResultsReporter.get_run()

Signed-off-by: Zack Cerza <zack@redhat.com>

suite: parse filter args into lists sooner

Signed-off-by: Zack Cerza <zack@redhat.com>

Merge pull request #951 from ceph/wip-shaman-2

Support fetching packages from shaman instead of gitbuilder

Reviewed-by: Dan Mick <dmick@redhat.com>