Cherry-pick notes:
- Options defined in src/common/options.cc in Octopus vs src/common/options/rgw.yaml.in
- RGWQuotaCache::get_stats does not take optional_yeild or DoutPrefixProvider arguments in Octopus
J. Eric Ivancich [Tue, 15 Jun 2021 19:20:33 +0000 (15:20 -0400)]
rgw: when deleted obj removed in versioned bucket, extra del-marker added
After initial checks are complete, this will read the OLH earlier than
previously to check the delete-marker flag and under the bug's
conditions will return -ENOENT rather than create a spurious delete
marker.
Cherry-pick notes:
- RGWRados::apply_olh_log does not take DoutPrefixProvider in Octopus
- change to use some namespace-qualified names in cls_rgw_types
Jeegn Chen [Wed, 25 Nov 2020 09:15:25 +0000 (17:15 +0800)]
rgw: avoid infinite loop when deleting a bucket
When deleting a bucket with an incomplete multipart upload that
has about 2000 parts uploaded, we noticed an infinite loop, which
stopped s3cmd from deleting the bucket forever.
Per check, when the bucket index was sharded (for example 128
shards), the original logic in
RGWRados::cls_bucket_list_unordered() did not calculate
the bucket shard ID correctly when the index key of a data
part was taken as the marker.
The issue is not necessarily reproduced each time. It will depend
on the key of the object. To reproduce it in 128-shard bucket,
we use 334 as the key for the incomplete multipart upload,
which will be located in Shard 127 (known by experiment). In this
setup, the original logic will usually come out a shard ID smaller
than 127 (since 127 is the largest one) from the marker and
thus a circle is constructed, which results in an infinite loop.
PS: Some times the bucket ID calculation may incorrectly going forward
instead of backward. Thus, the check logic may skip some shards,
which may have regular keys. In such scenarios, some non-empty buckets may
be deleted by accident.
In /etc/sysconfig/ceph we allow operators to define if ceph daemons
should be restarted on upgrade: CEPH_AUTO_RESTART_ON_UPGRADE.
But the post selinux scripts will stop ceph.target regardless if this
is set to `no`, leading to operators adding various hacks to prevent
these unexpected or inconvenient daemon restarts. By now, if users
are using rpms directly, they are likely orchestrating their own
daemon restarts so should not rely on the rpm itself to do this.
Fixes: https://tracker.ceph.com/issues/21672 Signed-off-by: Dan van der Ster <daniel.vanderster@cern.ch>
(cherry picked from commit 092a6e3e83e9ef8e37cb6f1033c345dcb5224cfc)
Igor Fedotov [Tue, 31 Aug 2021 12:54:23 +0000 (15:54 +0300)]
os/bluestore: fix bluefs migrate command
After migrating DB volume to a slow one RocksDB still
needs to be provided with slow.db path to properly access relevant files under db.slow subfolder.
Without that specification it tries to access them under 'db' one which
results in "not-found" error.
Fixes: https://tracker.ceph.com/issues/40434 Signed-off-by: Igor Fedotov <ifedotov@suse.com>
(cherry picked from commit 90852d9b6f0da7967121200c9a1c56bed1929d2d)
When running the `lvm migrate` subcommand without any args then the
ceph-volume command fails with a stack trace.
Traceback (most recent call last):
File "/usr/lib/python3.6/site-packages/ceph_volume/decorators.py", line 59, in newfunc
return f(*a, **kw)
File "/usr/lib/python3.6/site-packages/ceph_volume/main.py", line 151, in main
terminal.dispatch(self.mapper, subcommand_args)
File "/usr/lib/python3.6/site-packages/ceph_volume/terminal.py", line 194, in dispatch
instance.main()
File "/usr/lib/python3.6/site-packages/ceph_volume/devices/lvm/main.py", line 46, in main
terminal.dispatch(self.mapper, self.argv)
File "/usr/lib/python3.6/site-packages/ceph_volume/terminal.py", line 194, in dispatch
instance.main()
File "/usr/lib/python3.6/site-packages/ceph_volume/devices/lvm/migrate.py", line 520, in main
self.migrate_osd()
File "/usr/lib/python3.6/site-packages/ceph_volume/decorators.py", line 16, in is_root
return func(*a, **kw)
File "/usr/lib/python3.6/site-packages/ceph_volume/devices/lvm/migrate.py", line 403, in migrate_osd
if self.args.osd_id:
AttributeError: 'Migrate' object has no attribute 'args'
That's because we're exiting the parse_argv function but we continue to
execute the migrate_osd function. We should instead exit from the main function.
This update the parsing argument to have the same code than new-db and
new-wal classes.
Now the parsing is done in the make_parser function but the argv testing is
done in the main function allowing to exit the program and displaying the
help message when no arguments are provided.
The `ceph-volume lvm migrate/new-db/new-wal` commands don't support
running on non systemd systems or within containers.
Like other ceph-volume commands (lvm activate/batch/zap or raw activate)
we also need to be able to use the --no-systemd flag.
This is a regression introduced by 9212420, when the host is using a
logical partition then lsblk reports that partition as a child from the
physical device.
That logical partition is prefixed by the `└─` character.
This leads the `raw list` subcommand to show the lsblk error on the stderr.
```
$ ceph-volume raw list
{}
stderr: lsblk: `-/dev/sda1: not a block device
```
Igor Fedotov [Wed, 18 Aug 2021 10:39:02 +0000 (13:39 +0300)]
os/bluestore: accept undecodable multi-block bluefs transactions on log
replay.
We should proceed with OSD startup when detecting undecodable bluefs
transaction spanning multiple disk blocks during log replay.
The rationale is that such a transaction might appear during unexpected
power down - just not every disk block is written to disk. Hence we can
consider this a normal log replay stop condition.
libudev uses fnmatch(3) for matching attributes, meaning that shell
glob pattern matching is employed instead of literal string matching.
Escape glob metacharacters to suppress pattern matching.
Yin Congmin [Wed, 30 Jun 2021 08:56:23 +0000 (16:56 +0800)]
common/buffer: fix SIGABRT in rebuild_aligned_size_and_memory
There is such a bl, which needs to satisfy two conditions:
1)all ptrs' length sum except last ptr is aligned with 4K;
2)the length of last ptr is 0.
This bl will cause stack corruption when calling
bufferlist::rebuild_aligned_size_and_memory().
Deal with this special scenario in rebuild_aligned_size_and_memory() to
solve the bug. And added a specialtest-case to reproduce this scenario.
pybind/rbd: explain why "primary" isn't exposed in mirror_image_status_list()
"primary" is part of mirror image info (rbd_mirror_image_info_t) and
is exposed in mirror_image_get_status(). mirror_image_status_list(),
even though it is often thought of as an equivalent of repeated calls
to mirror_image_get_status(), doesn't actually fetch the mirror image
info.
pybind/rbd: actually append site_status dict to remote_statuses
Using += operator is wrong -- only site_status keys get appended
(and repeatedly at that in case there is more than one remote site
as the keys are added one by one).
Will Smith [Fri, 23 Jul 2021 19:18:12 +0000 (15:18 -0400)]
rbd: Fix mirror_image_get_status in rbd python bindings
When retrieving the status of a mirrored image from the Python rbd
library, a TypeError is raised.
*To Reproduce:*
Set up two Ceph clusters for block storage, and configure image
mirroring between their pools. Create a least one image with mirroring
enabled, then run the following script on either cluster (once the image
exists everywhere):
with rados.Rados(conffile=CONF_PATH) as cluster:
with cluster.open_ioctx(POOL_NAME) as ioctx:
with rbd.Image(ioctx, IMAGE_LABEL) as image:
image.mirror_image_get_status()
```
This will result in the following stack trace:
```
Traceback (most recent call last):
File "repo-bug.py", line 10, in <module>
image.mirror_image_get_status()
File "rbd.pyx", line 3363, in rbd.requires_not_closed.wrapper
File "rbd.pyx", line 5209, in rbd.Image.mirror_image_get_status
TypeError: list indices must be integers or slices, not str
```
Conflicts:
src/pybind/mgr/dashboard/controllers/rgw.py
- rgw-list endpoint doesn't have daemon_name parameter in octopus, so adopted
changes accordingly.
Deepika Upadhyay [Wed, 23 Jun 2021 05:12:38 +0000 (10:42 +0530)]
mon/PGMap: DIRTY field as N/A in `df detail` when cache tier not in use
'ceph df detail' reports a column for DIRTY objects under POOLS even
though cache tiers not being used. In replicated or EC pool all objects
in the pool are reported as logically DIRTY as they have never been
flushed .
we display N/A for DIRTY objects if the pool is not a cache tier.
See Rook issue https://github.com/rook/rook/issues/7940 for full
information.
Ceph bluestore disks can sometimes appear as though they have "phantom"
Atari (AHDI) partitions created on them when they don't in reality. This
is due to a series of bugs in the Linux kernel when it is built with
Atari support enabled. This behavior does not appear for raw mode OSDs on
partitions, only on disks.
Changing the on-disk format of Bluestore OSDs comes with
backwards-compatibility challenges, and fixing the issue in the Kernel
could be years before users get a fix. Working around the Kernel issue
in ceph-volume is therefore the best place to fix the issue for Ceph.
To work around the issue in Ceph volume, there are two behaviors that need
adjusted:
1. `ceph-volume inventory` should not report that a partition is
available if the parent device is a BlueStore OSD.
2. `ceph-volume raw list` should report parent disks if the disk is a
BlueStore OSD and not report the disk's children, BUT it should still
report children if the parent disk is not a BlueStore OSD.
Using only the exit status of `ceph-bluestore-tool show-label` to
determine if a device is a bluestore OSD could report a false negative
if there is a system error when `ceph-bluestore-tool` opens the device.
A better check is to open the device and read the bluestore device
label (the first 22 bytes of the device) to look for the bluestore
device signature ("bluestore block device"). If ceph-volume fails to
open the device due to a system error, it is safest to assume the device
is BlueStore so that an existing OSD isn't overwritten.
Tim Serong [Thu, 5 Aug 2021 08:48:57 +0000 (18:48 +1000)]
mgr/cephadm: pass --container-init to `cephadm deploy` if specified
This (sort of) reverts a small part of 442904bf2da, in order to
work correctly with 82d30afc413. The problem is that the latter
commit sets CONTAINER_INIT=False, which means we need to explicitly
pass --container-init to cephadm if we want that behaviour turned on.
There's a kink though: --container-init is only valid with cephadm's
"adopt", "bootstrap" and "deploy" commands; it's not a global option.
The cephadm mgr module doesn't invoke "adopt" or "bootstrap", only
"deploy", so we're only setting it in that specific case.
Fixes: https://tracker.ceph.com/issues/52064 Signed-off-by: Tim Serong <tserong@suse.com>
David Caro [Wed, 16 Jun 2021 08:32:15 +0000 (10:32 +0200)]
monitoring/grafana/cluster: use per-unit max and limit values
The value we get is a perunit, so the limits and the max value should
be over 1, not 100. Note that the value being shown was correct, it
was the gauge that was not showing the correct indicators.
projects / ceph.git / log