Keep pacific at high for now. Since teuthology-dispatcher locks nodes
before starting jobs, the upgrade jobs no longer need high priority to
run on more machines than average.
Signed-off-by: Josh Durgin <jdurgin@redhat.com>
Use the day of the year instead of compressing everything into
quarters. This lets us get through more breadth of combinations in a
year.
Signed-off-by: Josh Durgin <jdurgin@redhat.com>
smoke: 3x a week on master, weekly on other branches
pacific: every 2 days
master, octopus, nautilus, upgrades: weekly
This should be a net reduction in tests each week. Currently
we schedule far more than the lab can run in a given week.
Try to balance the scheduling over the course of the week.
Signed-off-by: Josh Durgin <jdurgin@redhat.com>
Include specifics about which pgs are affect, which pgs prevent us from
being ok to stop, etc.
The primary downside I see here is that a success and failure output will
look more similar to a human user
Signed-off-by: Sage Weil <sage@newdream.net>
* refs/pull/39138/head:
qa: valgrind test for cephfs-mirror daemon
cephfs-mirror: use preforker for daemonizing
test: adjust sleep time to account for valgrind runs
cephfs-mirror: gracefully shutdown threads, timers, etc..
cephfs-mirror: call ceph_release() to cleanup mount alloc
cephfs-mirror: shutdown filesystem/cluster connections on shutdown
cephfs-mirror: set init failed flag on FSMirror::init() failure
Reviewed-by: Patrick Donnelly <pdonnell@redhat.com>
To avoid a time window when m_update_status_task is invalid. If
during this time the cancel_update_mirror_image_replay_status is
called, it may cancel some other's ImageReplayer task, if it
happened to add the task with the same address.
Fixes: https://tracker.ceph.com/issues/49418
Signed-off-by: Mykola Golub <mgolub@suse.com>
`__all__` is only required if you want to modify the `form .. import *`
behavior. We don't need it.
Signed-off-by: Sebastian Wagner <sebastian.wagner@suse.com>
To cover the following case:
- Client A has image opened but does not owns the lock.
- Client B renames the image (client A is not aware of it).
- Client A becomes the lock owner.
- Client B requests rename, which is proxied to the client A.
Signed-off-by: Mykola Golub <mgolub@suse.com>
in general, `ActivePyModules::set_config()` is called by mgr module when
serving user commands updating module, sometimes if the option is of the
wrong type or invalid value, monitor rejects this request sent by mgr,
but the error info is only logged in the logging message on mgr, but not
returned to user. in this change, `ceph_set_module_option()` and the
underlying methods are updated to return the error to the caller as an
python exception.
Signed-off-by: Kefu Chai <kchai@redhat.com>
Add description for options --id and --client_fs to the ceph-fuse manual
and move description for -d closer to -f since both options are similar.
Signed-off-by: Rishabh Dave <ridave@redhat.com>
We have always assumed this to be the case and users' scripts and
orchestration tools have grown to depend on this. Let's add some
enforcement, prompted by [1]:
"I am running my Kubernetes worker node inside of an LXC container
which doesn't benefit from the device node created by the kernel, so
I'm using udev to create the /dev/rbd* device nodes inside of the LXC
container."
which, through the unfortunate interaction with ceph-csi rbd plugin,
results in data loss for "volumeMode: Filesystem" PVs because it ends
up recreating the filesystem every time the PV is attached to the pod:
"When deleting the pod and re-creating it, I can see that the RBD
image is indeed being reformatted. This seems to be because when
blkid is being run to check if the image is formatted, the /dev/rbd*
device has not yet been created by udev. By the time the code gets
down to running mkfs, the device is there and the damage is done."
[1] https://github.com/ceph/ceph-csi/issues/1820
Fixes: https://tracker.ceph.com/issues/49410
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
stringify(snap_id) is ephemeral:
tools/cephfs_mirror/PeerReplayer.cc:916:62: warning: object backing the
pointer will be destroyed at the end of the full-expression
snap_metadata snap_meta[] = {{PRIMARY_SNAP_ID_KEY.c_str(), stringify(snap_id).c_str()}};
Signed-off-by: Ronen Friedman <rfriedma@redhat.com>
Calculate the scaled cost for an OpSchedulerItem in msec based on the
overall osd capacity (across all shards). Remove logic to cache the
cost of the previous op to re-use for the next transaction as it doesn't
seem to provide any benefit in terms of performance.
Signed-off-by: Sridhar Seshasayee <sseshasa@redhat.com>
mclock clients are categorized into client, background_recovery and
background_best_effort. Each client category is required to specify
the mclock parameters like reservation, weight and limit. Therefore,
a simple structure (ClientAllocs) is introduced that encapsulates the
mclock parameters and is used to store those parameters for each client
category.
Other changes:
- The "high_client_ops" profile is now set as the default profile
instead of the "balanced" profile.
- Replace "Stringify" with std::to_string().
Signed-off-by: Sridhar Seshasayee <sseshasa@redhat.com>
* refs/pull/39328/head:
tools/rados: replace omap_key_valid with std::optional.
tools/rados: replace omap_key_pretty with prettify().
tools/rados: add support for binary object names.
Reviewed-by: J. Eric Ivancich <ivancich@redhat.com>
* refs/pull/39460/head:
msg/async/dpdk: be compatible with DPDK 19.11
common/Tub: remove Tub.h
cmake: use NUMA for package name in FindNUMA.cmake
msg/async/dpdk: define MARKER in .hh
msg/async/dpdk: use optional<> instead of Tub<>
compressor/lz4: use optional<> instead of Tub<>
compressor/lz4: use range-based loop
Reviewed-by: Josh Durgin <jdurgin@redhat.com>
