We weren't waiting for export dir to complete (the asok
just starts the process). This wasn't noticeable when running
remotely due to latency between the test runner and the MDS,
but it shows up when running against a local vstart cluster.
Signed-off-by: John Spray <john.spray@redhat.com>
I am seeing a strange thing where it seems like sometimes
a ls of /sys/fs/fuse/connections is returning empty when
connections do exist. It is pretty easy to make this
a non-issue by waiting for "more conns than we started with"
instead of "list of conns is different", so do that.
Signed-off-by: John Spray <john.spray@redhat.com>
Previously failure to stat mnt dir was interpreted
as being unmounted. For "transport endpoint no connected"
error we do want to recognise that it is mounted, albeit
with no ceph-fuse process.
Signed-off-by: John Spray <john.spray@redhat.com>
Use this during test setup to check whether
a filesystem is configured at all, before
trying to tear it down.
Signed-off-by: John Spray <john.spray@redhat.com>
So that my vstart subclass can put ./ before
all the commands.
One could set $PATH, but I like to unambiguously point
it at the local built binaries in case someone also
has some systemwide packages.
Signed-off-by: John Spray <john.spray@redhat.com>
A run failed due to thrashing.. missed by about 30s (the osd
eventually sent the last reply but we'd already timed out).
Signed-off-by: Sage Weil <sage@redhat.com>
The existing logic is to ceph-deploy osd create --zap-disk which will
zap the data device before preparing it. However it will not zap the
journal device (see http://tracker.ceph.com/issues/13291).
If ceph-deploy osd create fails, a fall back will zap both the data
device and the journal and try prepare again. This could work if
the device preparation and activation was synchronous and catch all
errors that could be caused by an unclean journal device. However,
the activation is asynchronous and it is entirely possible for a device
to be prepared successfully and fail to activate in the background.
The data and journal device are always zapped before calling ceph-deploy
osd create. The logic is simpler and the overhead is low.
http://tracker.ceph.com/issues/13000Fixes: #13000
Signed-off-by: Loic Dachary <loic@dachary.org>
CentOS 6.5 needs to install a package and reboot to grow the root file
system. Instead of assuming a common user-data.txt file can fit all
Operating Systems, make one user data per os-type/os-version combination.
Signed-off-by: Loic Dachary <loic@dachary.org>
The process run by flock must not inherit the file descriptor because
this will cause the lock to be held forever should the command survive
the call to flock. This is for instance the case for the ssh-agent.
Signed-off-by: Loic Dachary <loic@dachary.org>
Instead of relying on git_base_url, use the get_ceph_git_url() to obtain
the URL of the Ceph repository to use with git clone. This allows the
user to override it via the git_ceph_url configuration option and the
--git-ceph-url command line option to teuthology-openstack.
http://tracker.ceph.com/issues/11883 Refs: #11883
Signed-off-by: Loic Dachary <loic@dachary.org>
The config paramter of download_ceph_deploy does not have a ceph-deploy
item, therefore the ceph-deploy-branch parameter is always assumed to be
master.
Signed-off-by: Loic Dachary <loic@dachary.org>
Otherwise we can get
2015-09-24T19:22:15.191 INFO:teuthology.orchestra.run.mira080.stderr:Error ENXIO: problem getting command descriptions from osd.1
Signed-off-by: Sage Weil <sage@redhat.com>
This is the correct implementation of 685d76a77c,
merged while broken in ff1655cb57 and
reverted in 4cccde634f.
Signed-off-by: John Spray <john.spray@redhat.com>
This reverts commit ff1655cb57, reversing
changes made to 2b25080d4f.
Since we haven't actually started the MDS daemons yet, this code is broken.
Signed-off-by: Greg Farnum <gfarnum@redhat.com>
Our ffsb and fsync tests contain so many small writes at random offsets
that it can take >10 minutes to commit all of them to disk if we get
a slower OSD cluster. 15 minutes is still a plenty-fast timeout for
this stage compared to just hanging and losing the logs!
Signed-off-by: Greg Farnum <gfarnum@redhat.com>
Build Ceph packages from source for the required revision, os_type,
os_version and architecture and upload them to the gitbuilder
repository.
http://tracker.ceph.com/issues/13031Fixes: #13031
Signed-off-by: Loic Dachary <loic@dachary.org>
/var/run/ceph is 770. This is mainly necessary for any
interaction with the daemon sockets, but it is what users do
and it may avoid log noise.
Signed-off-by: Sage Weil <sage@redhat.com>
We need to be able to merge things into s3-tests master that
break rgw. Create ceph-foo branches (ceph-master,
ceph-infernalis, etc.) and use those instead.
Signed-off-by: Sage Weil <sage@redhat.com>
Previously were defaulting to a string, which
always compared greater than elapsed, so never
timed out.
Fixes: #12820
Signed-off-by: John Spray <john.spray@redhat.com>
When keep_running is true, do not shutdown the cluster, leave it as it
is for other workunits or tasks to use. This effectively allows the
ceph-deploy task to be used as a helper to deploy clusters.
The call to build_ceph_cluster is simplified by giving it the whole
configuration dictionary instead of re-building one with selected arguments.
Signed-off-by: Loic Dachary <loic@dachary.org>
When ceph-deploy fails, run ceph report to get more information about
the state of the cluster at the time of the failure.
Signed-off-by: Loic Dachary <loic@dachary.org>
In our RHCS 1.3 ceph-deploy docs, we tell users to run "ceph-deploy
install --cli" on their calamari admin node, but our smoke test wasn't
actually doing this.
See https://bugzilla.redhat.com/1252929 , "[Ubuntu 1.3.0] - ceph-deploy
install --no-adjust-repos --cli `hostname` is failing with a Traceback
error"
In RHCS 1.2 we don't have a /mnt/MON directory. The intention of
35c6363a1e was to handle this condition,
but in 1.2, the non-zero return code makes Teuthology fail the whole
test.
We don't want *Teuthology* itself to act on the return code here; we
simply want to know what it was and structure the rest of the test
accordingly.
lttng is not yet part of any private repo; since 1.3.0/CentOS is not
a supported product, just grab it from EPEL for this test.
Signed-off-by: Dan Mick <dan.mick@redhat.com>
Otherwise, ceph-deploy will install an apt source that points to
ceph.com, which will override the local ISO repos.
No --mon/--osd yet until 12147 is fixed
Signed-off-by: Dan Mick <dan.mick@redhat.com>
also waits to remove it from dead_osds. this fixes an issue where
do_sighup tries to send a signal to an osd that has not been revived
yet.
Signed-off-by: Andrew Schoen <aschoen@redhat.com>
This will call Thrasher.do_sighup which picks a random osd and sends a
signal.SIGHUP to it, delaying for the value of sighup_delay between each
time it picks a new osd to signal.
Signed-off-by: Andrew Schoen <aschoen@redhat.com>
This method runs in a separate greenlet than do_thrash and will pick a
random live osd to send a signal.SIGHUP to. There is a config option,
sighup_delay, which controls how long to delay between sending the
signals.
Signed-off-by: Andrew Schoen <aschoen@redhat.com>
