Ensure that the rados client binary doesn't segfault when specifying a
number of parameters without a corresponding --pool parameter.
Signed-off-by: David Disseldorp <ddiss@suse.de>
Because of a missing return, ceph-disk prepare would fail if given a
regular file as a journal. If the journal file does not exist, ceph-disk
will create it but fail to ensure that the ceph user owns it. The
symlink to the journal file is not set when the journal file is
specified on the command line and the journal file does not exist at
all. The ceph-osd daemon will silently create it as a file but it will
not be the file given in argument.
Add a test case to verify using a regular file as a journal works as
expected.
Fixes: http://tracker.ceph.com/issues/17662
Signed-off-by: Jayashree Candadai <jayaajay@indiana.edu>
Signed-off-by: Loic Dachary <ldachary@redhat.com>
The sh function will collect both stderr and stdout and debug
will mess the json parsing.
Fixes: http://tracker.ceph.com/issues/17607
Signed-off-by: Loic Dachary <ldachary@redhat.com>
After recently added image metadata replication it is not possible any
more to update it on non-primary image.
Signed-off-by: Mykola Golub <mgolub@mirantis.com>
wait_for_image_replay_stopped returns not when the state is stopped,
but when the state is not replaying. So a race was possible when an
asok command was running when the previos stop command was still in
progress, leading to unexpected results.
Signed-off-by: Mykola Golub <mgolub@mirantis.com>
By switching to a new gf-complete with SIMD runtime detection, we can now remove all the different flavors of jerasure and shec. This simplifies deployment and configuration of erasure coding, enables hetergenous OSDs, and enables us to take advantage of new performance improvements in jerasure without config/build changes.
This commit removes flavors from cmake, removes ErasureCodePluginSelect___, and fixes unit tests. There is now a single plugin for jerasure and a single plugin for shec.
SIMDExt.cmake was changed so that its a little more generic, and is not polluted with gf-complete specific CFLAG defines. The #define for SIMD instruction were based on gf-complete.
I also added a small init helper for jerasure that has code that was common between jerasure and shec.
Signed-off-by: Bassam Tabbara <bassam.tabbara@quantum.com>
this reverts d053705. i disabled this test in hope to bisect the
offending tests that fail the mysterious jenkins failure, which was
fixed by 6f3ce3a.
Signed-off-by: Kefu Chai <kchai@redhat.com>
The variable 'pgs_per_osd' set value from 'new_pgs' divided by 'expected_osds',
and its type is integer. So it would remove the decimal point and get smaller value.
This would have problem in some situations, for exmaple:
The limitation of pg creating for one OSD is '32'.
There have 3 OSDs and I want to increase pgs for a pool.
It should be the limitation for creating new pgs up to '96(32 * 3)' at once.
Now, I create '98' pgs for a pool.
In original code, '98' would be divided by 'expected_osds' and get the floating value '32....'
Because of the type which is integer, the 'pgs_per_osd' would be set to 32.
Then the value won't bigger than the limitation and get the wrong result.
Signed-off-by: DesmondS <desmond.s@inwinstack.com>
Fixes: http://tracker.ceph.com/issues/17169
Without a timeout on the command, it may hang for a very long time,
hunting for new mons. If it hangs for more than 60 seconds, it is
safe to assume the mon is indeed down.
Fixes: http://tracker.ceph.com/issues/16477
Signed-off-by: Loic Dachary <loic@dachary.org>
The scsi_debug SCSI devices do not have a symlink in /dev/disk/by-partuuid
because they are filtered out by 60-persistent-storage.rules. That was
worked around by 60-ceph-partuuid-workaround-rules which has been
removed by 9f76b9ff31.
Add create rules targetting this specific case, only for tests since the
problem does not show in real use cases.
Fixes: http://tracker.ceph.com/issues/17100
Signed-off-by: Loic Dachary <loic@dachary.org>
For newly created cluster the CEPH_OSDMAP_REQUIRE_KRAKEN will be
automatically set, while for existing clusters it will not.
This change add "require_jewel_osds" to white list, so user
can access it by the "ceph osd set *" command family.
Signed-off-by: xie xingguo <xie.xingguo@zte.com.cn>
Add ability to set monitor_interface and public_network values (as opposed
to using hard coded values)
This change should be squashed with the previous change before merging this
pull request
Signed-off-by: Warren Usui <wusui@magna002.ceph.redhat.com>
When a primiry image is being deleted, the mirrored image might
temporary be reported in error state, before deletion is propagated.
Signed-off-by: Mykola Golub <mgolub@mirantis.com>
- log to stderr;
- log status if a `wait_for` function failed;
- don't needlessly sleep in `wait_for` functions after the last
unsuccessful iteration;
- make `wait_for_pool_images` work for image removal case;
- fix `wait_for_pool_images` reset timeout (last_count set).
Signed-off-by: Mykola Golub <mgolub@mirantis.com>
We are seeing an issue due to the lockdep symbols
in libcephfs and librados clashing, which shows itself
after a fork in the flock tests. We can avoid this
by splitting the libcephfs tests that require librados
(access.cc) into their own compilation unit so that
the flock tests can run in a libcephfs-only process.
Fixes: http://tracker.ceph.com/issues/16556
Signed-off-by: John Spray <john.spray@redhat.com>
Snapshot rename operations utilize the (cluster) unique snapshot
sequence to prevent attempts at replays. When mirroring to a
different cluster, these sequences will not align.
Signed-off-by: Jason Dillaman <dillaman@redhat.com>
- This script is always called from a controlled environment
- use CEPH_BIN for exec's, otherwise QA sets PATH correctly
Signed-off-by: Willem Jan Withagen <wjw@digiware.nl>
create temp directory and files in $TMPDIR. the $TMPDIR is hard-wired to
/tmp before this change, we'd better respect the env variable $TMPDIR,
so it would be more consistent, and easier to do the cleanup if any.
Signed-off-by: Kefu Chai <kchai@redhat.com>
If an image is being bootstrapped, it implies that the rbd-mirror
daemon currently has the image open. The removal API will prevent the
removal of any image that is opened by another client.
Works-around: http://tracker.ceph.com/issues/16555
Signed-off-by: Mykola Golub <mgolub@mirantis.com>
this fixes failures like,
/home/jenkins-build/build/workspace/ceph-pull-requests/qa/workunits/cephtool/test.sh:
line 32: ceph osd blacklist ls | grep 192.168.0.1: command not found
where the failure is not the "failure" we are expecting.
in our tests, following command
expect_false "ceph osd blacklist ls | grep 192.168.0.1"
is designed to to verify that "ceph osd blacklist ls | grep 192.168.0.1"
fails with non-zero return code. but expect_false() evaluates the command
line using plain "$@", which will send the arguments direct to the shell,
and $0 is "ceph auth get client.xx | grep caps | grep mon", which does
not exist and is not built-in command. so we need to check the grep
command instead.
for multiple piped command line, use
expect_false sh <<< "echo foo | grep bar | grep baz"
Signed-off-by: Kefu Chai <kchai@redhat.com>
we set the CEPH_CLI_TEST_DUP_COMMAND enn var to verify the successful
commands are idempotent. but some of them are just not. among the other
things:
- ceph tell mds.a exit
- ceph tell mds.a respawn
the respawn command restart the mds daemon, its bind port changes and
all run-time status are reset. so strictly speaking, even the from the
point of view of client, this command is not idempotent. further more,
it fails the test, if the client sends the 2nd command too soon. because
the monitor might not able to update the re-spawned mds address before
the client asking for the new fsmap. so the cephfs client will just
use the old address of the specified mds, and hence will send the
request to port no one is listening anymore.
Signed-off-by: Kefu Chai <kchai@redhat.com>
This will help to test edge cases where the remote image does
or does not own the exclusive lock when the sync starts.
Signed-off-by: Jason Dillaman <dillaman@redhat.com>
When this test is failing and reach the limits, reading the log doesn't make
obvious that we reach them.
This simple patch adds the iterations numbers inside the output log.
Signed-off-by: Erwan Velu <erwan@redhat.com>
With the changes to ensure that the commit position of a new
client is initialized to the minimum position of other clients,
the 'journal inspect/export' commands return zero records because
the master client has committed all of its entries.
Workaround this by restoring the initial commit position after
writing to the image.
Fixes: http://tracker.ceph.com/issues/16011
Signed-off-by: Mykola Golub <mgolub@mirantis.com>
The debug messages from 'rbd bench-write' and 'rbd snap create',
in addition to the existing debug messages from rbd-mirror, make
it possible to determine the source of any image inconsistency.
Signed-off-by: Jason Dillaman <dillaman@redhat.com>
This test repeatedly runs rbd bench-write, kills the process
randomly to create an unclean journal shutdown, and verifies
that the image content replicates correctly.
Signed-off-by: Jason Dillaman <dillaman@redhat.com>
* test_osd_bench: the injectargs call actually fails due to the spaces
at the beginning. so remove the spaces in args before sending it to
injectargs
* update/add some injectargs tests accordingly
Signed-off-by: Kefu Chai <tchaikov@gmail.com>
A set of scripts used to bring up an OpenStack instance on one site,
connect a ceph cluster, add pools to store OpenStack glance images,
cinder volumes, nova vms, and cinder backups on rbd, and bring up a nova
compute node.
rbd_map_image() echoes the id of the image and is used in command
substitution:
id=$(rbd_map_image "${image}")
Output from rbd map isn't consumed and clobbers the return.
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
This is a followup to 38a572011d ("qa: rbd/map-snapshot-io.sh: don't
chown sysfs files") for the rest of the workunits:
- /sys/bus/rbd/add_single_major is used if present, so chown'ing
/sys/bus/rbd/add doesn't help. Use sudo on rbd map instead.
- Don't modprobe or udevadm settle - rbd CLI tool takes care of that.
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
/sys/bus/rbd/add_single_major is used if present, so chown'ing
/sys/bus/rbd/add doesn't help. Use sudo on rbd map instead.
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
As a precaution to using cleanup for mass deletion of other
objects, only allow --prefix which begins with "benchmark_data."
Signed-off-by: David Zafman <dzafman@redhat.com>
Initially the asok commands were registered only after the image
replayer start (and unregistered on stop) because their names were
built using remote pool and image names, which became known only after
start.
Now, the asok commands are registered on the image replayer construction
using the temporary name "remote_pool_name/global_image_id". They are
re-registered using "remote_pool_name/remote_image_name" when the
image replayer is started. Also the commands are not unregistered on
the image replayer stop.
Signed-off-by: Mykola Golub <mgolub@mirantis.com>
With these set, this script can work in teuthology, where we want to
control the how daemons run and their log locations, valgrind
settings, etc. The workunit task always sets CEPH_ID, and allows
other env vars to be configured optionally.
The RBD_MIRROR_USE_EXISTING_CLUSTER option can also speed up local testing.
Signed-off-by: Josh Durgin <jdurgin@redhat.com>
Daemon name is only useful to differentiate asok and pid filenames, so
just inline its use there. Everywhere else replace the 'daemon'
parameter with cluster or local_cluster as appropriate.
Remove the CLUSTER{1,2}_DAEMON vars as well, since they are no longer
used.
Signed-off-by: Josh Durgin <jdurgin@redhat.com>
"btrfs subvolume {list,delete}" needs root privilege even if the current
user owns this subvol. one can only list/delete he/she is the root, or
if the btrfs volume was mounted with "-o user_subvol_rm_allowed".
Signed-off-by: Kefu Chai <kchai@redhat.com>
in cf24535, we use $CEPH_ROOT to specify the $top_srcdir to unify
cmake and autotools, but this breaks ceph-qa-suite/tasks/workunit.py,
as it only clones the necessary qa/workunits directory, and does not
pass $CEPH_ROOT to the test scripts. so we need to set a default
$CEPH_ROOT if it is not set.
Signed-off-by: Kefu Chai <kchai@redhat.com>
Replaced relative paths in test/cephtool-test-mon.sh,
qa/workunits/cephtool/test.sh, and test/cephtool-test-mon.sh
to work with CEPH_FOO environment variables set in cmake.
Signed-off-by: Ali Maredia <amaredia@redhat.com>
Replaced relative paths in encode-decode-non-regression.sh
to work with CEPH_FOO environment variables set in
cmake.
Signed-off-by: Ali Maredia <amaredia@redhat.com>
Moved all the libraries in CMAKE_BINARY_DIR/lib
and all the binaries in CMAKE_BINARY_DIR/bin. Set
various environment variables for test-ceph-helpers.
Put those variables throughout
qa/workunits/ceph-helpers.sh.
NOTE: This is a very rough draft of these fixes.
Signed-off-by: Ali Maredia <amaredia@redhat.com>
FreeBSD once in a while forgets to remove *pid files (this is probably a bug).
But taking care of it this way is probably much in line of what is actually needs to be done
Signed-off-by: Willem Jan Withagen wjw@digiware.nl
Protect a number of unstable/experimental features behind durable flags
https://github.com/ceph/ceph/pull/8383
Reviewed-by: John Spray <john.spray@redhat.com>
As a precaution to using cleanup for mass deletion of other
objects, only allow --prefix which begins with "benchmark_data."
Signed-off-by: David Zafman <dzafman@redhat.com>
Method preprocess_remove_snaps() is designed to fast check whether
we can safely handle a remove-snaps-request without changing the osdmap.
The original design is to be able to handle snaps from multiple pools,
including those snaps even from a non-existent pool by simply skipping
over them. However, this method will quit on successfully detecting
any vaild snap which is truly needed to be removed and forward this
request to prepare_remove_snaps() for further processing.
From the above analysis, the prepare_remove_snaps() method will
theoretically also encounter some snaps which possibly belong to
non-existent pools.
This pr solves the above problem by adding a sanity check against
pool existense associated with the specified snap to be removed, which
shall be considered as a defensive move and makes prepare_remove_snaps()
stronger.
Signed-off-by: xie xingguo <xie.xingguo@zte.com.cn>
The current code was waiting 10s to expect the file being put.
If the file was put in a shorter time than 10s, the test just waits for
nothing reducing the execution speed of that test.
This patch simply check if the file is actually available every second
during 10sec to exit prematurely.
This patch saves exactly 10 sec on a local system, surely a little bit
less on an infra but still saves time.
Signed-off-by: Erwan Velu <erwan@redhat.com>
The actual code double the wait time between two calls leading to a
possible 511s of waiting time which sounds a little bit excessive.
This patch offer to reduce the global wait time to 300s and test more
often the rados status to exit the loop earlier. In a local test, that
saves 6 secs per run.
Signed-off-by: Erwan Velu <erwan@redhat.com>
ceph_watch_wait() is doing a sleep _before_ doing the test which could
stop this loop.
It's better doing the action first as it could exit immediately and
avoid a useless sleep.
That's a minor optimization but everything count when trying to get
something smooth.
Signed-off-by: Erwan Velu <erwan@redhat.com>
OSDs are taking some time to be up but waiting 10 secs seems execessive
here between two loops. In the worst case, we can be in a situation of
waiting 10secs for nothing as we are just a few microsecs after the osd
is up.
This patch simply reduce the sleep from 10 to 1 seconds.
Signed-off-by: Erwan Velu <erwan@redhat.com>
It could sounds like nothing but the actual sleeping rampup is counter
productive.
The code does : kill <proc>; sleep 0; kill <proc>; sleep 0; kill <proc;
sleep 1; and then it grows up 120 seconds by a smooth rampup.
But actually there is almost no chance the process dies so fast meaning
that by default we switch to the sleep 1.
Moving from sleep 0 to sleep 1 doesn't seems a big win but as
kill_daemons() is called very often we can save a lot of time by then
end.
This patch offer to sleep first a 1/10th of second instead of 0 and then
1/20th of second instead of 0.
The sleep call is also moved after the kill call as it's not necessary
waiting before executing the command.
This patch makes the running time of a test like osd-scrub-repair.sh
dropping from 7m30 to 7m7.
Saving another ~30seconds is an interesting win at make check level.
Signed-off-by: Erwan Velu <erwan@redhat.com>
wait_for_clean() is a very common call when running the make check.
It does wait the cluster to be stable before continuing.
This script was doing the same calls twice and could be optimized by
making the useful calls only once.
is_clean() function was checking num_pgs & get_num_active_clean()
The main loop itself was also calling get_num_active_clean()
This patch is inlining the is_clean() inside this loop to benefit from a
single get_num_active_clean() call. This avoid a useless call of (ceph +
xmlstarlet).
This patch does move all the 'timer reset' conditions into an else
avoiding spawning other ceph+xmlstarlet call while we already know we
should reset the timer.
The last modification is to reduce the sleeping time as the state of the
cluster is changing very fast.
This whole patch could looks like almost not a big win but for a test
like test/osd/osd-scrub-repair.sh, we drop from 9m56 to 9m30 while
reducing the number system calls.
At the scale of make check, that's a lot of saving.
Signed-off-by: Erwan Velu <erwan@redhat.com>
get_num_active_clean() is called very often but spawn 1 useless process.
The current "grep -v | wc -l" can be easily replaced by "grep -cv" which
do the same while spawning one process less.
Signed-off-by: Erwan Velu <erwan@redhat.com>
The current code of kill_daemons() was killing daemons one after the
other and wait it to actually die before switching to the next one.
This patch makes the kill_daemons() loop being run in parallel to avoid
this bottleneck.
Signed-off-by: Erwan Velu <erwan@redhat.com>
