Jason Dillaman [Thu, 11 Feb 2021 20:54:01 +0000 (15:54 -0500)]
librbd/mirror: leave non-primary snapshot images in creating state
The creating state is a special case in rbd-mirror where it will
automatically delete the image since it assumes it's malformed.
A non-primary, snapshot-based mirror image needs to have at least
one non-primary snapshot and the first one is not created until
after replay has started. Now rbd-mirror will update the mirror
image state to the enabled state after creating the first
non-primary snapshot but before attempting the sync.
Fixes: https://tracker.ceph.com/issues/49238 Signed-off-by: Jason Dillaman <dillaman@redhat.com>
(cherry picked from commit 43f2c208fa3042d93e4810d804ffe28e9ca7af77)
Jason Dillaman [Thu, 11 Feb 2021 20:45:01 +0000 (15:45 -0500)]
rbd-mirror: ensure that the last non-primary snapshot cannot be pruned
Tweak the normal pruning behavior to ensure that an incomplete initial
non-primary snapshot is not included in the prune set since we know
it will be complete since otherwise the image would have been deleted
due to not updating the mirror-image-state to enabled. Also ensure
we cannot prune a non-primary mirror snapshot if we don't have a
predecessor.
Jason Dillaman [Thu, 11 Feb 2021 20:23:56 +0000 (15:23 -0500)]
rbd-mirror: update snapshot mirror image state after snapshot creation
The non-primary mirror snapshot is what is used to link the non-primary
to the primary image. If there is an interruption between creating the
non-primary image and the creation of the first non-primary snapshot,
the images will be considerered unlinked.
A future commit will modify librbd to avoid setting the mirror image
state to enabled for non-primary snapshot-based mirroring images.
rbd-mirror will already automatically delete images in the CREATING
state during the bootstrap phase.
If the requested write length does not match the provided bufferlist
length, disable the move optimization and instead fallback to creating
a new sub-bufferlist for the object request.
Fixes: https://tracker.ceph.com/issues/49173 Signed-off-by: Jason Dillaman <dillaman@redhat.com>
(cherry picked from commit 8dbb4a3d971d9a48c171f161f531956dd0030403)
It appears that commit 6eb8f30a238 broke the test utility and
its failure was masked by the test case that expected a failure
due to a timeout force-killing the app.
Fixes: https://tracker.ceph.com/issues/49117 Signed-off-by: Jason Dillaman <dillaman@redhat.com>
(cherry picked from commit 8643b046fb4d5b05b4c75b83f16cd8ccc6a8b0a0)
Ilya Dryomov [Mon, 8 Feb 2021 16:01:47 +0000 (17:01 +0100)]
librbd: don't hold owner_lock for validate_image_removal()
handle_exclusive_lock() and handle_shut_down_exclusive_lock() call
validate_image_removal() without owner_lock held, so holding it in
shut_down_exclusive_lock() appears to be redundant.
Ilya Dryomov [Sun, 7 Feb 2021 14:09:24 +0000 (15:09 +0100)]
librbd: treat EROFS as expected in handle_acquire_lock()
If the peer refuses to release exclusive lock (e.g. in case automatic
exclusive lock transitions are disabled), EROFS is retured. Suppress
a rather confusing "Read-only file system" error message -- this case
is no different from EBUSY or EAGAIN.
Ilya Dryomov [Sun, 7 Feb 2021 12:46:15 +0000 (13:46 +0100)]
librbd: refuse to release exclusive lock when removing
Commit 25c2ffe145be ("librbd: acquire exclusive lock from peer when
removing") changed PreRemoveRequest to request exclusive lock from the
peer instead of giving up and proceeding without exclusive lock. This
caused one of the test cases that sometimes runs concurrent "rbd rm"
against the same image to fail intermittently, most often on assert
because exclusive lock is now automatically transitioned to another
"rbd rm" on its request.
The root cause is older and probably goes back to when synchronous
librbd::remove() which held owner_lock across all operations including
trim_image() was converted to a set of state machines. Since then, any
peer that requests exclusive lock (instead of trying once and backing
off) is able to mess with image removal.
Install StandardPolicy to disable automatic exclusive lock transitions
during image removal.
Jason Dillaman [Mon, 8 Feb 2021 16:53:28 +0000 (11:53 -0500)]
rbd-mirror: don't prune older mirror snapshots when pruning incomplete snapshot
Since we normally prune in order, we need to ensure that we don't prune older
snapshots when we need to delete an incomplete mirror snapshot since the
older snapshot might be the only remaining mirror snapshot.
Jason Dillaman [Fri, 29 Jan 2021 15:44:38 +0000 (10:44 -0500)]
librbd/deep_copy: skip snap list if object is known to be clean
If the fast-diff indicates that the destination object should exist
and that it hasn't changed, there shouldn't be a need to issue the
snap list operation. Instead, just update the destination object map
to indicate the existence of the object.
Jason Dillaman [Fri, 29 Jan 2021 02:42:09 +0000 (21:42 -0500)]
librbd/deep_copy: object-copy state machine must update object map
If there was no data to copy, the object-copy state machine was bypassing
the object-map update states and prematurely completing. Since the
object-map is default-initialized to all non-existent objects, this results
in incorrect state for OBJECT_EXISTS_CLEAN objects.
Jason Dillaman [Wed, 3 Feb 2021 18:21:34 +0000 (13:21 -0500)]
librbd/io: track object non-existence when computing snapshot deltas
Re-use the existing DNE state to track whether or not the object
already exists when computing snapshot deltas from an arbitrary
set of snapshots. Previously, the non-existence of the object was
only computed for snap id 0 for tracking whiteouts. In a future
commit, the deep-copy object-copy state machine will be able to
properly update the object-map state to indicate exists clean
vs non-existent state.
Jason Dillaman [Wed, 3 Feb 2021 15:13:28 +0000 (10:13 -0500)]
librbd/io: only track initial diff extents if no diffs exists
The purpose of the initial diff extents ({0, 0}) was to help track
whether or not objects exists for read-from-parent / whiteout
tracking. Once we have at least one set of diffs on the object, we
actually have enough information to know about the object state.
Jason Dillaman [Thu, 28 Jan 2021 23:30:16 +0000 (18:30 -0500)]
librbd/object_map: diff state machine should track object existence
The deep-copy snapshot-create state machine initializes the object-map
state to non-existent for all objects. There was an assumption that the
deep-copy object-copy state machine would always update the object map
but that was being skipped for clean objects as an optimization. This
change will support a future commit to run the object-copy state machine
for existing objects.
Lucian Petrut [Wed, 3 Feb 2021 08:59:24 +0000 (08:59 +0000)]
win32*.sh: move debug symbols to separate files
This patch simplifies releasing Windows binaries along with debug
symbols.
By default, we're going to provide minimum debug information (-g1).
The symbols are extracted from the binaries and placed in separate
files in the ".debug" folder, which is used by gdb implicitly.
This is more convenient than having separate versions of the binaries,
with or without debug symbols.
Lucian Petrut [Fri, 29 Jan 2021 11:03:20 +0000 (11:03 +0000)]
rbd: propagate WNBD start errors
This change will propagate the errors that WNBD may return when
spinning up the IO workers.
Also, we'll avoid removing the registry record for failed
non-persistent mappings. Those will be cleaned up when the service
restarts or when explicitly unmapped.
Lucian Petrut [Fri, 29 Jan 2021 09:54:10 +0000 (09:54 +0000)]
rbd: improve Windows remap failure handling
At the moment, if an image can't be remapped when the centralized
RBD service starts, the service will stop and already started
daemons will continue running.
This change adds a new option: "--remap-failure-fatal". If set,
when an image can't be remmaped, the service stops AND cleans up
the running daemons. By default, an error will be logged and the
service will continue running.
Lucian Petrut [Thu, 28 Jan 2021 14:08:29 +0000 (14:08 +0000)]
rbd: add image map timeouts on Windows
This change adds configurable timeouts used when starting the centralized RBD
service and mapping images.
This change also fixes an issue where the service would wait indefinitely for
a failed mapping. This issue was caused by the fact that multiple child
processes were inheriting pipe handles.
Note that we can't use timeouts with Windows anonymous pipes, which is why
we're going to use unique named pipes.
Lucian Petrut [Wed, 27 Jan 2021 12:58:48 +0000 (12:58 +0000)]
rbd: restrict Windows service exec calls
The centralized Ceph Windows service is responsible of managing
rbd-wnbd daemons. When starting, it's respawns the daemons using the
command line saved in the Windows registry. Also, for new mappings,
the command line is passed through a named pipe.
While writing to the according named pipe and windows registry entries
requires admin privileges, it's better to avoid running arbitrary
commands.
This patch will drop the executable from the commands that the
Ceph service accepts. Instead, it will only accept arguments that
are passed to the the binary that was used to start the service
(rbd-wnbd.exe).
Lucian Petrut [Wed, 27 Jan 2021 09:06:19 +0000 (09:06 +0000)]
rbd: allow non persistent Windows mappings
At the moment, all Windows RBD mappings are persistent, being
recreated when the Ceph service starts.
This change adds the "--non-persistent" flag to allow skipping
certain images.
Note that even for non-persistent mappings, we're still storing
data in the Windows registry, allowing us to retrieve image
details such as the rbd pool as well as the admin socket path.
When the daemons stop or the Ceph service starts, non-persistent
entries are cleaned up from the Windows registry.
Lucian Petrut [Tue, 26 Jan 2021 14:13:22 +0000 (14:13 +0000)]
rbd: delegate map requests to the Windows service
At the moment, rbd-wnbd daemons are tied to the Windows
session. When the user logs off or the WinRM remote session
terminates, the daemons are killed.
In order to avoid such issues, we'll delegate the daemon
initialization to the centralized Ceph Windows service.
We're using a named pipe for the Ceph service communication.
We're not using Ceph admin sockets for now since unix sockets
aren't available on Windows Server 2016.
Worth mentioning that the Ceph Windows service will restart
registered mappings when the host reboots based on the command
lines saved in the Windows registries. Writing to the registry
key as well as the service named pipe requires admin privileges.
Venky Shankar [Tue, 18 Aug 2020 09:32:38 +0000 (05:32 -0400)]
test: optionally create a backup filesystem on startup
Also filter out client-id's starting with "mirror" when
cleaning leftover auth-ids since teuthology would be
configured to create client.mirror and client.mirror_remote
clients before executing mirroring tests.
Ronen Friedman [Thu, 28 Jan 2021 17:04:35 +0000 (19:04 +0200)]
osd: fix the scrubber behavior on multiple preemption attempts
Latest scrub code creates a time window in which a specific scrub
is marked as "preempted", but future preemptions are prohibited.
Write operations handled are then blocked but not restarted on time.
Yuri Weinstein [Wed, 3 Feb 2021 18:14:56 +0000 (10:14 -0800)]
Merge pull request #39165 from sebastian-philipp/pacific-backport-39035-38935-39028-39087-38998-38854-38982-39101-39043-39113-
pacific: cephadm: Batch backport January (2)
Reviewed-by: Jan Fajerski <jfajerski@suse.com> Reviewed-by: Neha Ojha <nojha@redhat.com> Reviewed-by: Laura Paduano <lpaduano@suse.com> Reviewed-by: Juan Miguel Olmo Martínez <jolmomar@redhat.com>
Jan Fajerski [Tue, 22 Dec 2020 13:29:02 +0000 (14:29 +0100)]
ceph-volume: add some flexibility to bytes_to_extents
For batch we want to report the projected sizes before creating any LVM
structures, we use the device size for that. This means we pass this
projected size to lvm/api.py::create_lv. This could result in situations
where we request more extents for the new LV than are actually
available, as LVM takes up some available space on the block device
(plus alignment).
We fix this by adjusting the extent count downwards if its less than 1% off the
requested value.
Fixes: https://tracker.ceph.com/issues/47758 Signed-off-by: Jan Fajerski <jfajerski@suse.com>
(cherry picked from commit ece6e7eb9a92cb944b2f201a05a5acfb877c2dba)
Neha Ojha [Tue, 2 Feb 2021 17:43:56 +0000 (17:43 +0000)]
qa/*/test_envlibrados_for_rocksdb: use osd_client_message_cap to prevent slow requests
EnvLibradosMutipoolTest.DBBulkLoadKeysInRandomOrder can overload OSDs and cause
heartbeat timeouts. Tests in test_envlibrados_for_rocksdb also generate slow
requests on OSDs. Use osd_client_message_cap to prevent this.
Since this option is disabled by default, this may be a good way to exercise it.
Kamoltat [Thu, 7 Jan 2021 15:39:19 +0000 (15:39 +0000)]
mgr/pg_autoscaler: avoid scale-down until there is pressure
The autoscaler will start out with scaling each
pools to have a full complements of pgs from the start
and will only decrease it when pools need more due to
increased usage.
Introduced a unit test that tests only the
function get_final_pg_target_and_ratio() which
deals with the distrubtion of pgs amongst the
pools
Edited workunit script to reflect the change
of how pgs are calculated and distrubted.
Ilya Dryomov [Sun, 31 Jan 2021 14:24:32 +0000 (15:24 +0100)]
qa/tasks/rbd: don't wait for krbd symlink to be created/removed
Commit 3754c665a11e (":qa/tasks/rbd: test qemu on top of rbd
encryption") broke dev_create() for krbd by messing up the "wait
for the symlink to be created by udev" loop. The rbd tool has been
synchronizing with udev internally since 2014, so rather than fixing
let's just drop it.
Ilya Dryomov [Sun, 31 Jan 2021 13:27:52 +0000 (14:27 +0100)]
qa/tasks/rbd: dev_create() expects a properties dict
In order to add encryption_format property, commit 3754c665a11e
(":qa/tasks/rbd: test qemu on top of rbd encryption") changed the
dev_create() contract. It now expects a properties dict, but the
main task routine wasn't updated and still passes role_images.
On top of that, role_images itself got broken. It is supposed
to be a mapping from role to a corresponding image name, not to
a dict with an image_name key. This affected generic_mount().
Sage Weil [Fri, 29 Jan 2021 20:03:43 +0000 (14:03 -0600)]
qa/tasks/cephadm: allow custom git_url for cephadm_branch pull
Normally the git_url is git://git.ceph.com/ceph-ci.git, which mirrors
upstream ceph-ci.git. However, the release branches aren't present there.
Allow a custom git_url so we can pull these from the main ceph.git.
Kefu Chai [Thu, 28 Jan 2021 12:35:36 +0000 (20:35 +0800)]
cmake: install rook-client-python using ExternalProject
so we don't need to rerun the generate_rook_ceph_client.sh script
everytime when building the script. cmake creates a stamp file for
tracking the dependencies and the time of modification of dependencies.
Sage Weil [Wed, 27 Jan 2021 21:44:21 +0000 (15:44 -0600)]
python-common: fix test_datetime_to_str_2 on non-UTC hosts
The old test parsed to a datetime without a tz, which was interpreted as
the local time zone when rendering back to a string. Specify that it's a
UTC datetime so that behavior is consistent regardless of the test host
timezone.
Paul Cuzner [Wed, 27 Jan 2021 20:40:58 +0000 (09:40 +1300)]
cephadm:add missing kernel_security property
The propery decorator had gone missing which meant
the dump of host facts was missing the kernel security
(LSM) settings. This patch just adds the @property
decorator back
Kefu Chai [Sun, 24 Jan 2021 06:58:51 +0000 (14:58 +0800)]
cephadm: refactor call() using asyncio.asyncio.StreamReader
simpler this way, also fix a couple issues:
* create a child watcher explicitly, see
https://bugs.python.org/issue35621
* use StringIO for collecting outputs for better performance,
instead of appending the lines to an existing str
* catch ValueError when reading from the stream reader,
because StreamReader.readline() could raise ValueError when
it reaches the buffer limit while looking for a separator.
in this case, we should try again, in hope that the spawned
process can feed the reader with more data which contains
the separator (i.e., b'\n').
* backport ThreadedChildWatcher from Python 3.8 so we can
run create_subprocess_exec() in non-main threads.
iSCSI daemons need to execute <ceph service status> command
Fixes: https://tracker.ceph.com/issues/48925 Signed-off-by: Juan Miguel Olmo Martínez <jolmomar@redhat.com>
(cherry picked from commit 9b9934f75b648a9ee01848710a6f7150a371e26e)
Adam King [Mon, 11 Jan 2021 18:23:04 +0000 (13:23 -0500)]
mgr/cephadm: force flag for ok-to-stop and ok-to-stop for monitoring stack
Daemons that could cause data loss when stopped will always block.
Daemons that will only cause loss in availability should block but have
a workaround in the form of a force flag if the user is okay with the service
being down.
Also implements ok-to-stop for monitoring stack daemons that uses this system
of blocking on availability loss unless force flag is provided
Signed-off-by: Adam King <adking@redhat.com> Signed-off-by: Daniel-Pivonka <dpivonka@redhat.com>
(cherry picked from commit aeeffb07f1bb226b64a362de4eecd7d63fc0b0d1)
projects / ceph.git / log