ceph/PendingReleaseNotes

372 lines
22 KiB
Plaintext
Raw Blame History

This file contains ambiguous Unicode characters

This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

>=19.0.0
* The cephfs-shell utility is now packaged for RHEL 9 / CentOS 9 as required
python dependencies are now available in EPEL9.
* RGW: S3 multipart uploads using Server-Side Encryption now replicate correctly in
multi-site. Previously, the replicas of such objects were corrupted on decryption.
A new tool, ``radosgw-admin bucket resync encrypted multipart``, can be used to
identify these original multipart uploads. The ``LastModified`` timestamp of any
identified object is incremented by 1ns to cause peer zones to replicate it again.
For multi-site deployments that make any use of Server-Side Encryption, we
recommended running this command against every bucket in every zone after all
zones have upgraded.
* CEPHFS: MDS evicts clients which are not advancing their request tids which causes
a large buildup of session metadata resulting in the MDS going read-only due to
the RADOS operation exceeding the size threshold. `mds_session_metadata_threshold`
config controls the maximum size that a (encoded) session metadata can grow.
* CephFS: For clusters with multiple CephFS file systems, all the snap-schedule
commands now expect the '--fs' argument.
* CephFS: The period specifier ``m`` now implies minutes and the period specifier
``M`` now implies months. This has been made consistent with the rest
of the system.
* RGW: New tools have been added to radosgw-admin for identifying and
correcting issues with versioned bucket indexes. Historical bugs with the
versioned bucket index transaction workflow made it possible for the index
to accumulate extraneous "book-keeping" olh entries and plain placeholder
entries. In some specific scenarios where clients made concurrent requests
referencing the same object key, it was likely that a lot of extra index
entries would accumulate. When a significant number of these entries are
present in a single bucket index shard, they can cause high bucket listing
latencies and lifecycle processing failures. To check whether a versioned
bucket has unnecessary olh entries, users can now run ``radosgw-admin
bucket check olh``. If the ``--fix`` flag is used, the extra entries will
be safely removed. A distinct issue from the one described thus far, it is
also possible that some versioned buckets are maintaining extra unlinked
objects that are not listable from the S3/ Swift APIs. These extra objects
are typically a result of PUT requests that exited abnormally, in the middle
of a bucket index transaction - so the client would not have received a
successful response. Bugs in prior releases made these unlinked objects easy
to reproduce with any PUT request that was made on a bucket that was actively
resharding. Besides the extra space that these hidden, unlinked objects
consume, there can be another side effect in certain scenarios, caused by
the nature of the failure mode that produced them, where a client of a bucket
that was a victim of this bug may find the object associated with the key to
be in an inconsistent state. To check whether a versioned bucket has unlinked
entries, users can now run ``radosgw-admin bucket check unlinked``. If the
``--fix`` flag is used, the unlinked objects will be safely removed. Finally,
a third issue made it possible for versioned bucket index stats to be
accounted inaccurately. The tooling for recalculating versioned bucket stats
also had a bug, and was not previously capable of fixing these inaccuracies.
This release resolves those issues and users can now expect that the existing
``radosgw-admin bucket check`` command will produce correct results. We
recommend that users with versioned buckets, especially those that existed
on prior releases, use these new tools to check whether their buckets are
affected and to clean them up accordingly.
* CephFS: Running the command "ceph fs authorize" for an existing entity now
upgrades the entity's capabilities instead of printing an error. It can now
also change read/write permissions in a capability that the entity already
holds. If the capability passed by user is same as one of the capabilities
that the entity already holds, idempotency is maintained.
* CephFS: Two FS names can now be swapped, optionally along with their IDs,
using "ceph fs swap" command. The function of this API is to facilitate
file system swaps for disaster recovery. In particular, it avoids situations
where a named file system is temporarily missing which would prompt a higher
level storage operator (like Rook) to recreate the missing file system.
See https://docs.ceph.com/en/latest/cephfs/administration/#file-systems
docs for more information.
* CephFS: Before running the command "ceph fs rename", the filesystem to be
renamed must be offline and the config "refuse_client_session" must be set
for it. The config "refuse_client_session" can be removed/unset and
filesystem can be online after the rename operation is complete.
* RADOS: A POOL_APP_NOT_ENABLED health warning will now be reported if
the application is not enabled for the pool irrespective of whether
the pool is in use or not. Always tag a pool with an application
using ``ceph osd pool application enable`` command to avoid reporting
of POOL_APP_NOT_ENABLED health warning for that pool.
The user might temporarily mute this warning using
``ceph health mute POOL_APP_NOT_ENABLED``.
CephFS: Disallow delegating preallocated inode ranges to clients. Config
`mds_client_delegate_inos_pct` defaults to 0 which disables async dirops
in the kclient.
* S3 Get/HeadObject now support query parameter `partNumber` to read a specific
part of a completed multipart upload.
* RGW: Fixed a S3 Object Lock bug with PutObjectRetention requests that specify
a RetainUntilDate after the year 2106. This date was truncated to 32 bits when
stored, so a much earlier date was used for object lock enforcement. This does
not effect PutBucketObjectLockConfiguration where a duration is given in Days.
The RetainUntilDate encoding is fixed for new PutObjectRetention requests, but
cannot repair the dates of existing object locks. Such objects can be identified
with a HeadObject request based on the x-amz-object-lock-retain-until-date
response header.
* RADOS: `get_pool_is_selfmanaged_snaps_mode` C++ API has been deprecated
due to being prone to false negative results. It's safer replacement is
`pool_is_in_selfmanaged_snaps_mode`.
* RADOS: For bug 62338 (https://tracker.ceph.com/issues/62338), we did not choose
to condition the fix on a server flag in order to simplify backporting. As
a result, in rare cases it may be possible for a PG to flip between two acting
sets while an upgrade to a version with the fix is in progress. If you observe
this behavior, you should be able to work around it by completing the upgrade or
by disabling async recovery by setting osd_async_recovery_min_cost to a very
large value on all OSDs until the upgrade is complete:
``ceph config set osd osd_async_recovery_min_cost 1099511627776``
* RADOS: A detailed version of the `balancer status` CLI command in the balancer
module is now available. Users may run `ceph balancer status detail` to see more
details about which PGs were updated in the balancer's last optimization.
See https://docs.ceph.com/en/latest/rados/operations/balancer/ for more information.
* CephFS: Full support for subvolumes and subvolume groups is now available
for snap_schedule Manager module.
* RBD: When diffing against the beginning of time (`fromsnapname == NULL`) in
fast-diff mode (`whole_object == true` with `fast-diff` image feature enabled
and valid), diff-iterate is now guaranteed to execute locally if exclusive
lock is available. This brings a dramatic performance improvement for QEMU
live disk synchronization and backup use cases.
* RBD: The ``try-netlink`` mapping option for rbd-nbd has become the default
and is now deprecated. If the NBD netlink interface is not supported by the
kernel, then the mapping is retried using the legacy ioctl interface.
* RADOS: Read balancing may now be managed automatically via the balancer
manager module. Users may choose between two new modes: ``upmap-read``, which
offers upmap and read optimization simultaneously, or ``read``, which may be used
to only optimize reads. For more detailed information see https://docs.ceph.com/en/latest/rados/operations/read-balancer/#online-optimization.
* CephFS: MDS log trimming is now driven by a separate thread which tries to
trim the log every second (`mds_log_trim_upkeep_interval` config). Also,
a couple of configs govern how much time the MDS spends in trimming its
logs. These configs are `mds_log_trim_threshold` and `mds_log_trim_decay_rate`.
>=18.0.0
* The RGW policy parser now rejects unknown principals by default. If you are
mirroring policies between RGW and AWS, you may wish to set
"rgw policy reject invalid principals" to "false". This affects only newly set
policies, not policies that are already in place.
* The CephFS automatic metadata load (sometimes called "default") balancer is
now disabled by default. The new file system flag `balance_automate`
can be used to toggle it on or off. It can be enabled or disabled via
`ceph fs set <fs_name> balance_automate <bool>`.
* RGW's default backend for `rgw_enable_ops_log` changed from RADOS to file.
The default value of `rgw_ops_log_rados` is now false, and `rgw_ops_log_file_path`
defaults to "/var/log/ceph/ops-log-$cluster-$name.log".
* The SPDK backend for BlueStore is now able to connect to an NVMeoF target.
Please note that this is not an officially supported feature.
* RGW's pubsub interface now returns boolean fields using bool. Before this change,
`/topics/<topic-name>` returns "stored_secret" and "persistent" using a string
of "true" or "false" with quotes around them. After this change, these fields
are returned without quotes so they can be decoded as boolean values in JSON.
The same applies to the `is_truncated` field returned by `/subscriptions/<sub-name>`.
* RGW's response of `Action=GetTopicAttributes&TopicArn=<topic-arn>` REST API now
returns `HasStoredSecret` and `Persistent` as boolean in the JSON string
encoded in `Attributes/EndPoint`.
* All boolean fields previously rendered as string by `rgw-admin` command when
the JSON format is used are now rendered as boolean. If your scripts/tools
relies on this behavior, please update them accordingly. The impacted field names
are:
* absolute
* add
* admin
* appendable
* bucket_key_enabled
* delete_marker
* exists
* has_bucket_info
* high_precision_time
* index
* is_master
* is_prefix
* is_truncated
* linked
* log_meta
* log_op
* pending_removal
* read_only
* retain_head_object
* rule_exist
* start_with_full_sync
* sync_from_all
* syncstopped
* system
* truncated
* user_stats_sync
* RGW: The beast frontend's HTTP access log line uses a new debug_rgw_access
configurable. This has the same defaults as debug_rgw, but can now be controlled
independently.
* RBD: The semantics of compare-and-write C++ API (`Image::compare_and_write`
and `Image::aio_compare_and_write` methods) now match those of C API. Both
compare and write steps operate only on `len` bytes even if the respective
buffers are larger. The previous behavior of comparing up to the size of
the compare buffer was prone to subtle breakage upon straddling a stripe
unit boundary.
* RBD: compare-and-write operation is no longer limited to 512-byte sectors.
Assuming proper alignment, it now allows operating on stripe units (4M by
default).
* RBD: New `rbd_aio_compare_and_writev` API method to support scatter/gather
on both compare and write buffers. This compliments existing `rbd_aio_readv`
and `rbd_aio_writev` methods.
* The 'AT_NO_ATTR_SYNC' macro is deprecated, please use the standard 'AT_STATX_DONT_SYNC'
macro. The 'AT_NO_ATTR_SYNC' macro will be removed in the future.
* Trimming of PGLog dups is now controlled by the size instead of the version.
This fixes the PGLog inflation issue that was happening when the on-line
(in OSD) trimming got jammed after a PG split operation. Also, a new off-line
mechanism has been added: `ceph-objectstore-tool` got `trim-pg-log-dups` op
that targets situations where OSD is unable to boot due to those inflated dups.
If that is the case, in OSD logs the "You can be hit by THE DUPS BUG" warning
will be visible.
Relevant tracker: https://tracker.ceph.com/issues/53729
* RBD: `rbd device unmap` command gained `--namespace` option. Support for
namespaces was added to RBD in Nautilus 14.2.0 and it has been possible to
map and unmap images in namespaces using the `image-spec` syntax since then
but the corresponding option available in most other commands was missing.
* RGW: Compression is now supported for objects uploaded with Server-Side Encryption.
When both are enabled, compression is applied before encryption. Earlier releases
of multisite do not replicate such objects correctly, so all zones must upgrade to
Reef before enabling the `compress-encrypted` zonegroup feature: see
https://docs.ceph.com/en/reef/radosgw/multisite/#zone-features and note the
security considerations.
* RGW: the "pubsub" functionality for storing bucket notifications inside Ceph
is removed. Together with it, the "pubsub" zone should not be used anymore.
The REST operations, as well as radosgw-admin commands for manipulating
subscriptions, as well as fetching and acking the notifications are removed
as well.
In case that the endpoint to which the notifications are sent maybe down or
disconnected, it is recommended to use persistent notifications to guarantee
the delivery of the notifications. In case the system that consumes the
notifications needs to pull them (instead of the notifications be pushed
to it), an external message bus (e.g. rabbitmq, Kafka) should be used for
that purpose.
* RGW: The serialized format of notification and topics has changed, so that
new/updated topics will be unreadable by old RGWs. We recommend completing
the RGW upgrades before creating or modifying any notification topics.
* RBD: Trailing newline in passphrase files (`<passphrase-file>` argument in
`rbd encryption format` command and `--encryption-passphrase-file` option
in other commands) is no longer stripped.
* RBD: Support for layered client-side encryption is added. Cloned images
can now be encrypted each with its own encryption format and passphrase,
potentially different from that of the parent image. The efficient
copy-on-write semantics intrinsic to unformatted (regular) cloned images
are retained.
* CEPHFS: Rename the `mds_max_retries_on_remount_failure` option to
`client_max_retries_on_remount_failure` and move it from mds.yaml.in to
mds-client.yaml.in because this option was only used by MDS client from its
birth.
* The `perf dump` and `perf schema` commands are deprecated in favor of new
`counter dump` and `counter schema` commands. These new commands add support
for labeled perf counters and also emit existing unlabeled perf counters. Some
unlabeled perf counters became labeled in this release, with more to follow in
future releases; such converted perf counters are no longer emitted by the
`perf dump` and `perf schema` commands.
* `ceph mgr dump` command now outputs `last_failure_osd_epoch` and
`active_clients` fields at the top level. Previously, these fields were
output under `always_on_modules` field.
* `ceph mgr dump` command now displays the name of the mgr module that
registered a RADOS client in the `name` field added to elements of the
`active_clients` array. Previously, only the address of a module's RADOS
client was shown in the `active_clients` array.
* RBD: All rbd-mirror daemon perf counters became labeled and as such are now
emitted only by the new `counter dump` and `counter schema` commands. As part
of the conversion, many also got renamed to better disambiguate journal-based
and snapshot-based mirroring.
* RBD: list-watchers C++ API (`Image::list_watchers`) now clears the passed
`std::list` before potentially appending to it, aligning with the semantics
of the corresponding C API (`rbd_watchers_list`).
* The rados python binding is now able to process (opt-in) omap keys as bytes
objects. This enables interacting with RADOS omap keys that are not decodeable as
UTF-8 strings.
* Telemetry: Users who are opted-in to telemetry can also opt-in to
participating in a leaderboard in the telemetry public
dashboards (https://telemetry-public.ceph.com/). Users can now also add a
description of the cluster to publicly appear in the leaderboard.
For more details, see:
https://docs.ceph.com/en/latest/mgr/telemetry/#leaderboard
See a sample report with `ceph telemetry preview`.
Opt-in to telemetry with `ceph telemetry on`.
Opt-in to the leaderboard with
`ceph config set mgr mgr/telemetry/leaderboard true`.
Add leaderboard description with:
`ceph config set mgr mgr/telemetry/leaderboard_description Cluster description`.
* CEPHFS: After recovering a Ceph File System post following the disaster recovery
procedure, the recovered files under `lost+found` directory can now be deleted.
* core: cache-tiering is now deprecated.
* mClock Scheduler: The mClock scheduler (default scheduler in Quincy) has
undergone significant usability and design improvements to address the slow
backfill issue. Some important changes are:
* The 'balanced' profile is set as the default mClock profile because it
represents a compromise between prioritizing client IO or recovery IO. Users
can then choose either the 'high_client_ops' profile to prioritize client IO
or the 'high_recovery_ops' profile to prioritize recovery IO.
* QoS parameters like reservation and limit are now specified in terms of a
fraction (range: 0.0 to 1.0) of the OSD's IOPS capacity.
* The cost parameters (osd_mclock_cost_per_io_usec_* and
osd_mclock_cost_per_byte_usec_*) have been removed. The cost of an operation
is now determined using the random IOPS and maximum sequential bandwidth
capability of the OSD's underlying device.
* Degraded object recovery is given higher priority when compared to misplaced
object recovery because degraded objects present a data safety issue not
present with objects that are merely misplaced. Therefore, backfilling
operations with the 'balanced' and 'high_client_ops' mClock profiles may
progress slower than what was seen with the 'WeightedPriorityQueue' (WPQ)
scheduler.
* The QoS allocations in all the mClock profiles are optimized based on the above
fixes and enhancements.
* For more detailed information see:
https://docs.ceph.com/en/latest/rados/configuration/mclock-config-ref/
* mgr/snap_schedule: The snap-schedule mgr module now retains one less snapshot
than the number mentioned against the config tunable `mds_max_snaps_per_dir`
so that a new snapshot can be created and retained during the next schedule
run.
* `ceph config dump --format <json|xml>` output will display the localized
option names instead of its normalized version. For e.g.,
"mgr/prometheus/x/server_port" will be displayed instead of
"mgr/prometheus/server_port". This matches the output of the non pretty-print
formatted version of the command.
>=17.2.1
* The "BlueStore zero block detection" feature (first introduced to Quincy in
https://github.com/ceph/ceph/pull/43337) has been turned off by default with a
new global configuration called `bluestore_zero_block_detection`. This feature,
intended for large-scale synthetic testing, does not interact well with some RBD
and CephFS features. Any side effects experienced in previous Quincy versions
would no longer occur, provided that the configuration remains set to false.
Relevant tracker: https://tracker.ceph.com/issues/55521
* telemetry: Added new Rook metrics to the 'basic' channel to report Rook's
version, Kubernetes version, node metrics, etc.
See a sample report with `ceph telemetry preview`.
Opt-in with `ceph telemetry on`.
For more details, see:
https://docs.ceph.com/en/latest/mgr/telemetry/
* OSD: The issue of high CPU utilization during recovery/backfill operations
has been fixed. For more details, see: https://tracker.ceph.com/issues/56530.
>=15.2.17
* OSD: Octopus modified the SnapMapper key format from
<LEGACY_MAPPING_PREFIX><snapid>_<shardid>_<hobject_t::to_str()>
to
<MAPPING_PREFIX><pool>_<snapid>_<shardid>_<hobject_t::to_str()>
When this change was introduced, 94ebe0e also introduced a conversion
with a crucial bug which essentially destroyed legacy keys by mapping them
to
<MAPPING_PREFIX><poolid>_<snapid>_
without the object-unique suffix. The conversion is fixed in this release.
Relevant tracker: https://tracker.ceph.com/issues/56147
* Cephadm may now be configured to carry out CephFS MDS upgrades without
reducing ``max_mds`` to 1. Previously, Cephadm would reduce ``max_mds`` to 1 to
avoid having two active MDS modifying on-disk structures with new versions,
communicating cross-version-incompatible messages, or other potential
incompatibilities. This could be disruptive for large-scale CephFS deployments
because the cluster cannot easily reduce active MDS daemons to 1.
NOTE: Staggered upgrade of the mons/mgrs may be necessary to take advantage
of the feature, refer this link on how to perform it:
https://docs.ceph.com/en/quincy/cephadm/upgrade/#staggered-upgrade
Relevant tracker: https://tracker.ceph.com/issues/55715
* Introduced a new file system flag `refuse_client_session` that can be set using the
`fs set` command. This flag allows blocking any incoming session
request from client(s). This can be useful during some recovery situations
where it's desirable to bring MDS up but have no client workload.
Relevant tracker: https://tracker.ceph.com/issues/57090
* New MDSMap field `max_xattr_size` which can be set using the `fs set` command.
This MDSMap field allows to configure the maximum size allowed for the full
key/value set for a filesystem extended attributes. It effectively replaces
the old per-MDS `max_xattr_pairs_size` setting, which is now dropped.
Relevant tracker: https://tracker.ceph.com/issues/55725
* Introduced a new file system flag `refuse_standby_for_another_fs` that can be
set using the `fs set` command. This flag prevents using a standby for another
file system (join_fs = X) when standby for the current filesystem is not available.
Relevant tracker: https://tracker.ceph.com/issues/61599