Merge PR #65879 into wip-vshankar-testing-20260219.125903

* refs/pull/65879/head:

Reviewed-by: Venky Shankar <vshankar@redhat.com>

Merge PR #65881 into wip-vshankar-testing-20260219.125903

* refs/pull/65881/head:

Reviewed-by: Dhairya Parmar <dparmar@redhat.com>
Reviewed-by: Venky Shankar <vshankar@redhat.com>

Merge PR #65883 into wip-vshankar-testing-20260219.125903

* refs/pull/65883/head:

Reviewed-by: Venky Shankar <vshankar@redhat.com>

Merge PR #65885 into wip-vshankar-testing-20260219.125903

* refs/pull/65885/head:

Reviewed-by: Venky Shankar <vshankar@redhat.com>
Reviewed-by: Jos Collin <jcollin@redhat.com>

Merge PR #66288 into wip-vshankar-testing-20260219.125903

* refs/pull/66288/head:

Reviewed-by: Venky Shankar <vshankar@redhat.com>

Merge pull request #67221 from guits/node-proxy-various-fixes

node-proxy: major refactor and various fixes

Merge pull request #67277 from afreen23/nvmeof-api

mgr/dashboard: Add apis for add/del hosts on namespaces

Reviewed-by: Nizamudeen A <nia@redhat.com>

Merge pull request #67165 from Matan-B/wip-matanb-io_uring

src/CMakeLists.txt: Allow Seastar to reuse HAVE_LIBURING

Signed-off-by: Mohit Agrawal <moagrawa@redhat.com>

Merge pull request #67375 from shraddhaag/wip-shraddhaag-availability-default-state

doc: update default availbility score status

Merge pull request #67132 from rhcs-dashboard/delete-gateway-nodes

mgr/dashboard: delete-gateway-nodes

Merge pull request #67119 from ceph/copilot/add-copilot-instructions-file

github: define contribution workflows for regular and backport PRs

copilot: add GitHub Copilot instructions

Github allows to add a instructions file to each repo
(.github/copilot-instructions.md) to improve the behavior
of Copilot Reviews and Agent.

These instructions can also be customized per path, filetype, etc.:
https://docs.github.com/en/copilot/how-tos/configure-custom-instructions/add-repository-instructions

This commit was authored through a Github Agent session: https://github.com/ceph/ceph/tasks/edeca07b-eabd-477c-917a-a18e72a0e2c2

Co-authored-by: GitHub Copilot noreply@github.com
Generated-by: Claude Sonnet 4.5
Signed-off-by: Ernesto Puerta <epuertat@redhat.com>

Merge pull request #66527 from gardran/wip-gardran-dump-omap

os/bluestore: add omap_bytes perf counter.

Reviewed-by: Adam Kupczyk <akupczyk@ibm.com>

mgr/cephadm: validate hostname in NodeProxyCache

This adds a _resolve_hosts() method to resolve hostname from kwargs
and raise OrchestratorError when the host has no node-proxy data.

Fixes: https://tracker.ceph.com/issues/74749
Signed-off-by: Guillaume Abrioux <gabrioux@ibm.com>

node-proxy: improve HTTP error logging in client

This commit makes it log the http error with the code and the reason
in sessionservice_discover() and log the error code along with the
body in query() for 5xx responses.

Fixes: https://tracker.ceph.com/issues/74749
Signed-off-by: Guillaume Abrioux <gabrioux@ibm.com>

node-proxy: get serial number instead of SKU

Let's get the serial number instead of SKU.

Fixes: https://tracker.ceph.com/issues/74749
Signed-off-by: Guillaume Abrioux <gabrioux@ibm.com>

node-proxy: allow multiple sources per component

COMPONENT_SPECS can now be a single spec or a list of specs per component.
Data from all sources is merged. Unavailable paths are skipped.

Extract get_component_data() from update_component() to support the merge logic.

Fixes: https://tracker.ceph.com/issues/74749
Signed-off-by: Guillaume Abrioux <gabrioux@ibm.com>

node-proxy: re-auth and retry once on 401

This commit makes node-proxy clear the session, call login() and retry
the request when a 401 http error is caught.

Fixes: https://tracker.ceph.com/issues/74749
Signed-off-by: Guillaume Abrioux <gabrioux@ibm.com>

node-proxy: fix flake8 E721 in _dict_diff

Use "is not" instead of "!=" for type comparison in _dict_diff()

Fixes: https://tracker.ceph.com/issues/74749
Signed-off-by: Guillaume Abrioux <gabrioux@ibm.com>

node-proxy: make the update loop interval configurable

Read system.refresh_interval from config and use it in the update loop
sleep. The default value is 180s when unset.

Fixes: https://tracker.ceph.com/issues/74749
Signed-off-by: Guillaume Abrioux <gabrioux@ibm.com>

mgr/node-proxy: fix "ceph orch hardware status --category criticals"

The criticals path was using the wrong data shape:
node-proxy sends status as:

  component -> sys_id -> member

 but the code assumed:

  sys_id -> component -> member

This fixes get_critical_from_host() and _criticals_table() to iterate
in the correct order and build the criticals result with the right
nesting.

Fixes: https://tracker.ceph.com/issues/74749
Signed-off-by: Guillaume Abrioux <gabrioux@ibm.com>

node-proxy: normalize storage data per member

Let's apply normalize_dict() to each member's data only, so the first
level keys (that are redfish member identifiers like "Self") are not
lowercased.

This avoids duplicate entries in hardware status.

Example:

```
[root@node-proxy-1 cephadm]# ./cephadm shell -- ceph orch hardware status --category criticals
Inferring fsid 9d6d6012-067a-11f1-8e61-525400a04a72
Inferring config /var/lib/ceph/9d6d6012-067a-11f1-8e61-525400a04a72/mon.node-proxy-1/config
+--------------+-----------+------+--------+-------+
|     HOST     | COMPONENT | NAME | STATUS | STATE |
+--------------+-----------+------+--------+-------+
| node-proxy-1 |    self   | None |  N/A   |  N/A  |
| node-proxy-1 |    Self   | None |  N/A   |  N/A  |
+--------------+-----------+------+--------+-------+
```

Fixes: https://tracker.ceph.com/issues/74749
Signed-off-by: Guillaume Abrioux <gabrioux@ibm.com>

node-proxy: encapsulate send logic in dedicated method

Move the "send data to mgr when inventory changed" logic from main()
into a dedicated method _try_send_update().
This flattens the reporter loop and keeps main() to a single call under
the lock.

Fixes: https://tracker.ceph.com/issues/74749
Signed-off-by: Guillaume Abrioux <gabrioux@ibm.com>

node-proxy: log actual data delta in reporter

this adds a _dict_diff() function that computes recursive dict diff
and uses it in reporter to log the delta (truncated at 2048 chars)

Fixes: https://tracker.ceph.com/issues/74749
Signed-off-by: Guillaume Abrioux <gabrioux@ibm.com>

node-proxy: add periodic heartbeats in main and reporter loops

This logs an info message every 5 minutes so that logs show the agent
and reporter are still running when nothing else is logged.

Fixes: https://tracker.ceph.com/issues/74749
Signed-off-by: Guillaume Abrioux <gabrioux@ibm.com>

node-proxy: adjust log levels

Let's adjust log levels across the project:

- use warning for bad request in the API,
  when thread is not alive and for retry failure,
- use error for OOB load failure,
- use info for backoff interval,
- use debug in send attempts and for member fetch

Signed-off-by: Guillaume Abrioux <gabrioux@ibm.com>

node-proxy: add unit tests

This adds some unit tests.

Fixes: https://tracker.ceph.com/issues/74749
Signed-off-by: Guillaume Abrioux <gabrioux@ibm.com>

node-proxy: add tox config for mypy, flake8, isort, black

this adds tox.ini with environments to run mypy, flake8, isort, and
black on the ceph_node_proxy code.

Fixes: https://tracker.ceph.com/issues/74749
Signed-off-by: Guillaume Abrioux <gabrioux@ibm.com>

node-proxy: black and isort formatting pass

Format the code with black and isort.

Fixes: https://tracker.ceph.com/issues/74749
Signed-off-by: Guillaume Abrioux <gabrioux@ibm.com>

node-proxy: fix mypy errors

this commit fixes mypy errors by adding explicit types for get_path
and get_* getters methods, extending SystemBackend with
start/shutdown and declaring _ca_temp_file on NodeProxyManager

Fixes: https://tracker.ceph.com/issues/74749
Signed-off-by: Guillaume Abrioux <gabrioux@ibm.com>

node-proxy: handle nested Redfish paths for components

Add a _resolve_path() helper to support components whose data
lives under nested Redfish paths when assembling component data.

For instance, power is exposed at 'PowerSubsystem/PowerSupplies'

Fixes: https://tracker.ceph.com/issues/74749
Signed-off-by: Guillaume Abrioux <gabrioux@ibm.com>

node-proxy: split out config, bootstrap and redfish logic

refactor config, bootstrap, redfish layer, and monitoring:

this:
- adds a config module (CephadmCofnig, load_cephadm_config and
get_node_proxy_config) and protocols for api/reporter.
- extracts redfish logic to redfish.py
- adds a vendor registry with entrypoints.
- simplifies main() and NodeProxyManager().

Fixes: https://tracker.ceph.com/issues/74749
Signed-off-by: Guillaume Abrioux <gabrioux@ibm.com>

node-proxy: refactor config loading

This commit renames CONFIG to DEFAULTS and add load_config() with
deep merge, refactor Config to use path + defaults and makes
node-proxy config path configurable via bootstrap JSON or env.

Fixes: https://tracker.ceph.com/issues/74749
Signed-off-by: Guillaume Abrioux <gabrioux@ibm.com>

node-proxy: add 'vendor based' redfish system selection

This commit adds REDFISH_SYSTEM_CLASSES registry (generic, dell, ...),
this way the user can choose a system class.

The default value is BaseRedfishSystem (generic) when vendor isn't specified.

Fixes: https://tracker.ceph.com/issues/74749
Signed-off-by: Guillaume Abrioux <gabrioux@ibm.com>

node-proxy: introduce component spec registry and overrides for updates

This change introduces a single COMPONENT_SPECS dict and get_update_spec(component)
as the single source of truth for RedFish component update config (collection, path,
fields, attribute). To support hardware that uses different paths or attributes,
get_component_spec_overrides() allows overriding only those fields (via dataclasses.replace())
without duplicating the rest of the spec.
All _update_network, _update_power, etc. now call _run_update(component).

For instance, AtollonSystem uses this to set the power path to 'PowerSubsystem'.

Fixes: https://tracker.ceph.com/issues/74749
Signed-off-by: Guillaume Abrioux <gabrioux@ibm.com>

mgr/cephadm: safe status/health access in node-proxy agent and inventory

This adds helpers in NodeProxyEndpoint and NodeProxyCache to safely
read status.health and status.state.

In NodeProxyEndpoint, methods _get_health_value() and _get_state_value()
are used in get_nok_members() to avoid KeyError on malformed data.

In NodeProxyCache, _get_health_value(), _has_health_value(),
_is_error_status(), and _is_unknown_status() are used in fullreport()
and when filtering 'non ok' members instead of accessing
status['status']['health'] inline.

Fixes: https://tracker.ceph.com/issues/74749
Signed-off-by: Guillaume Abrioux <gabrioux@ibm.com>

node-proxy: narrow build_data exception handling and re-raise

With this commit, it catches only KeyError, TypeError, and
AttributeError in build_data() instead of Exception, and
re-raise after logging so callers get the actual error.

Fixes: https://tracker.ceph.com/issues/74749
Signed-off-by: Guillaume Abrioux <gabrioux@ibm.com>

node-proxy: refactor Endpoint/EndpointMgr and fix chassis paths

This commit refactors EndpointMgr and Endpoint to use explicit dicts
instead of dynamic attributes. It also fixes member path filtering
so chassis endpoints use Chassis paths.

Fixes: https://tracker.ceph.com/issues/74749
Signed-off-by: Guillaume Abrioux <gabrioux@ibm.com>

node-proxy: use safe field access in storage update

Replace direct dictionary access with .get() method when processing
storage fields to handle missing optional fields gracefully.

(extra change: extract get_members_names() call for better readability.)

Fixes: https://tracker.ceph.com/issues/74749
Signed-off-by: Guillaume Abrioux <gabrioux@ibm.com>

node-proxy: reduce log verbosity for missing optional fields

Change missing field logging from warning to debug level in
RedfishDellSystem, as missing optional fields can be expected behavior
and and doesn't require warning level logging.

Fixes: https://tracker.ceph.com/issues/74749
Signed-off-by: Guillaume Abrioux <gabrioux@ibm.com>

Merge pull request #66544 from NitzanMordhai/wip-nitzan-encoder-test-backward-incompability-checks

test/encoding/readable: Add backward incompat checks

Merge pull request #67287 from nbalacha/wip-nbalacha-71390

rgw/configstore:  don't reinitialize the rados client in RadosRealmWatcher

Reviewed-by: Casey Bodley <cbodley@redhat.com>
Reviewed-by: Adam Emerson <aemerson@redhat.com>

Merge pull request #66455 from cfanz/wip-cfanz-fix-rgw-counter-overflow

rgw: fix overflow of outstanding counter in SimpleThrottler

Reviewed-by: Casey Bodley <cbodley@redhat.com>

Merge pull request #67306 from cbodley/wip-rgw-rest-client-strftime

rgw/multisite: use libfmt to format Date header

Reviewed-by: Adam C. Emerson <aemerson@redhat.com>
Reviewed-by: Matt Benjamin <mbenjamin@redhat.com>

mgr/dashboard: Add apis for add/del hosts on namespaces

- these are UI APIs
- also removed namespace API and using "*" in existing instead for getting all ns in subsystem

Signed-off-by: Afreen Misbah <afreen@ibm.com>

Merge pull request #67112 from rhcs-dashboard/cephfs-module-enable

mgr/dashboard: add CephFS Mirroring enablement page

Reviewed-by: Afreen Misbah <afreen@ibm.com>

Merge pull request #66484 from samarahu/d4n-remove-erase-from-update

Reviewed-by: Pritha Srivastava <prsrivas@redhat.com>

doc: update default availbility score status

The feature is turned off by default post
https://github.com/ceph/ceph/pull/65545. Update the docs
to reflect the same.

Signed-off-by: Shraddha Agrawal <shraddha.agrawal000@gmail.com>

mgr/dashboard: delete-gateway-nodes

Fixes: https://tracker.ceph.com/issues/74336
Signed-off-by: Sagar Gopale <sagar.gopale@ibm.com>

Merge pull request #67348 from VallariAg/wip-nvmeof-udisks-disable

qa: Fix coredumps caused by udisks

doc/crimson/crimson.rst: introduce Enabling io_uring

We could enable this as part of packaged installs though
letting users enable this explicitly seems like a better approach.

Signed-off-by: Matan Breizman <mbreizma@redhat.com>

seastar: update submodule to wip-matanb-seastar-feb-26

```
Allow provided liburing builds via the URING::uring target.
This allows seastar, when built as a submodule, to use external
liburing.
```

https://github.com/ceph/seastar/tree/wip-matanb-seastar-feb26

Signed-off-by: Matan Breizman <mbreizma@redhat.com>

Merge pull request #67370 from kotreshhr/qa-mirror-flake8-fix

qa/test_mirroring: Fix flake8 errors

Reviewed-by: Ilya Dryomov <idryomov@gmail.com>
Reviewed-by: Venky Shankar <vshankar@redhat.com>

qa/test_mirroring: Fix flake8 errors

Introduced-by: c1e827247bd20e8a1851bc2d7a9861c12d033ef0
Signed-off-by: Kotresh HR <khiremat@redhat.com>

Merge pull request #67231 from guits/ceph-volume-inventory-ls-all

ceph-volume: include LVM mapper devices in get_devices()

Merge pull request #67110 from FredNass/patch-2

doc/radosgw: rgw_lua_max_memory_per_state defaults to 128K (not 500K)

Merge pull request #66967 from rhcs-dashboard/gateway-add-modal

mgr/dashboard: gateway-add-modal

Reviewed-by: Afreen Misbah <afreen@ibm.com>

Merge pull request #67305 from kotreshhr/qa-mirror

qa: Add retry logic to remove most sleeps in mirroring tests

Reviewed-by: Venky Shankar <vshankar@redhat.com>

Merge pull request #66912 from idryomov/wip-74394

osd/PrimaryLogPG: encode an empty data_bl for empty sparse reads

Merge pull request #66894 from tchaikov/wip-ec-isa-fix-cache-collision

erasure-code/isa: fix cache collision causing buffer overflow

Merge pull request #66376 from NitzanMordhai/wip-nitzan-self-test-influx-set-hostname

qa/tasks/mgr: test_module_selftest set influx hostname to avoid warnings

Merge pull request #62067 from ljflores/wip-tracker-67179

osd: add pg-upmap-primary to clean_pg_upmaps

Merge pull request #67353 from idryomov/wip-daemonwatchdog-unbound

qa/tasks/daemonwatchdog: fix unbound variable in bark_reason message

Reviewed-by: Patrick Donnelly <pdonnell@ibm.com>

Merge pull request #67351 from idryomov/wip-74712

qa: krbd_rxbounce.sh: do more reads to generate more errors

Reviewed-by: Ramana Raja <rraja@redhat.com>

Merge pull request #65318 from tchaikov/wip-mgr-progress-cleanup

pybind/mgr/progress: cleanups

Reviewed-by: Samuel Just <sjust@redhat.com>

qa/tasks/daemonwatchdog: fix unbound variable in bark_reason message

Introduced in commit 783f0e3a9903 ("qa: Adding a new class for the
daemonwatchdog to monitor").

Signed-off-by: Ilya Dryomov <idryomov@gmail.com>

qa: krbd_rxbounce.sh: do more reads to generate more errors

On faster hardware having each thread do 1024 reads isn't always
sufficient for the "two orders of magnitude" threshold that is used in
the test.

Fixes: https://tracker.ceph.com/issues/74712
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>

Merge pull request #67218 from jamiepryde/jerasure-unit-test-big-endian-fix

test/erasure-code: Use memory comparisons that are safe on big endian arch

qa: Fix coredumps caused by udisks

Also redo commands to install nvme 2.13.
Fixes: https://tracker.ceph.com/issues/74922
Signed-off-by: Vallari Agrawal <vallari.agrawal@ibm.com>

mgr/dashboard: add CephFS Mirroring enablement page

Fixes: https://tracker.ceph.com/issues/74633
Signed-off-by: Pedro Gonzalez Gomez <pegonzal@ibm.com>

Merge pull request #67240 from guits/cv-raw-list-perf

ceph-volume: single lvs call to speed up exclude_lvm_osd_devices

Merge pull request #66646 from ujjawal-ibm/ceph-volume-format-device

Add NVMe preformat support to skip redundant mkfs discard

Merge PR #66183 into main

* refs/pull/66183/head:

Reviewed-by: Anoop C S <anoopcs@cryptolab.net>
Reviewed-by: Shachar Sharon <ssharon@redhat.com>

Merge pull request #66606 from shraddhaag/add-logs

PeeringState::needs_recovery(): log first missing item

Merge pull request #66352 from tchaikov/debian-invoke-rc.d

debian: remove invoke-rc.d calls from postrm scripts

Merge pull request #66591 from NitzanMordhai/wip-nitzan-pickup-object-corpus-20.2.0-380-gdbcbbd3f281

ceph-object-corpus: pick up 20.2.0-380-gdbcbbd3f281

Merge pull request #66553 from smanjara/wip-fix-object-sync-enoent

rgw: retry ENOENT if bucket reshard races with index operation

Merge pull request #66721 from swjtu-zhanglei/main-wip

rgw: refact by encapsulating io_context_pool to enforce lazy initialization

Reviewed-by: Adam Emerson <aemerson@redhat.com>

Merge pull request #66567 from mertsunacoglu/wip-prerequest-auth

RGW: Change prerequest hook to run after authorization process

Reviewed-by: Yuval Lifshitz <ylifshit@redhat.com>

Merge pull request #66509 from morphes1995/rgw_qos_fix

rgw: ops large budget caused by rate limiter was enabled and max_ops …

Reviewed-by: Mark Kogan <mkogan@ibm.com>

qa: Fix test_cephfs_mirror_remote_snap_corrupt_fails_synced_snapshot

Fixes: https://tracker.ceph.com/issues/74896
Signed-off-by: Kotresh HR <khiremat@redhat.com>

qa: Add retry logic to remove most sleeps in mirroring tests

The mirroring tests contain lot of sleeps adding it up to ~1hr.
This patch adds a retry logic and removes most of them. This
is cleaner and saves considerable time in test time for mirroring.

Fixes: https://tracker.ceph.com/issues/74878
Signed-off-by: Kotresh HR <khiremat@redhat.com>

ceph-volume: skip redundant NVMe mkfs discards

- Avoid redundant discard during mkfs when discard is disabled
- Reduces mkfs time on large NVMe devices by skipping long running discard operations

Fixes: https://tracker.ceph.com/issues/74908
Signed-off-by: Ujjawal Anand <ujjawal.anand@ibm.com>

Merge pull request #67264 from guits/fix-74818

ceph-volume: pass osd_type value to ceph-volume

Merge pull request #66594 from cbodley/wip-qa-rgw-account-migration

qa/rgw: add test_account_migration workunit

Reviewed-by: Krunal Chheda <kchheda3@bloomberg.net>

Merge PR #66657 into main

* refs/pull/66657/head:
developer: Add name to maps

Reviewed-by: Shweta Bhosale <Shweta.Bhosale1@ibm.com>
Reviewed-by: Patrick Donnelly <pdonnell@ibm.com>

Merge PR #67309 into main

* refs/pull/67309/head:
doc: correct spelling

Reviewed-by: Zac Dover <zac.dover@gmail.com>
Reviewed-by: Patrick Donnelly <pdonnell@ibm.com>

Merge PR #67078 into main

* refs/pull/67078/head:
doc: Fixes a spelling error on the RATIO for PG scaling recommendation.

Reviewed-by: Anthony D Atri <anthony.datri@gmail.com>
Reviewed-by: Patrick Donnelly <pdonnell@ibm.com>

Merge pull request #66526 from gardran/wip-gardran-show-sharding

os/bluestore: show RocksDB sharding information

Reviewed-by: Adam Kupczyk <akupczyk@ibm.com>

Merge PR #67320 into main

* refs/pull/67320/head:
.github/workflows: use main branch for config diff

Reviewed-by: Ernesto Puerta <epuertat@redhat.com>
Reviewed-by: Naveen Naidu <naveen.naidu@ibm.com>

Merge pull request #66821 from rhcs-dashboard/cephfs-mirroring-filesystem

mgr/dashboard: Cephfs mirroring Filesystem Selection

Reviewed-by: Naman Munet <nmunet@redhat.com>
Reviewed-by: Pedro Gonzalez Gomez <pegonzal@redhat.com>

common/options/crimson.yaml.in: allow seastar to select available backend

* If non backend is selected, let seastar choose the default
  reactor backend (io_uring, if available).
* If a backend is selected, don't implicitly fallback to diffrent one.

Updated the conf option and moved to "Reactor options" section.

Signed-off-by: Matan Breizman <mbreizma@redhat.com>

cmake/modules/Builduring.cmake: add URING::uring as alias

Allow seastar submodule which expects URING:uring target name
to be imported.

Signed-off-by: Matan Breizman <mbreizma@redhat.com>

src/CMakeLists.txt: Set Seastar_IO_URING based on HAVE_LIBURING

Signed-off-by: Matan Breizman <mbreizma@redhat.com>

CMakeLists.txt: Generalize WITH_*_LIBURING

set WITH_LIBURING if HAVE_LIBAIO without depending on BLUESTORE.

Signed-off-by: Matan Breizman <mbreizma@redhat.com>

CMakeLists.txt: find aio for CRIMSON or BLUESTORE

aio is used to set WITH_LIBURING, instead of finding it twice try to
find aio if either flags are set.
Does not change existing behavior for Bluestore.

Signed-off-by: Matan Breizman <mbreizma@redhat.com>

CMakeLists.txt: Move WITH_CRIMON before setting WITH_LIBURING

Signed-off-by: Matan Breizman <mbreizma@redhat.com>

vstart.sh: fix identation

Signed-off-by: Matan Breizman <mbreizma@redhat.com>

crimson: Support io_uring as reactor backend option

Provide a bootable option to configure reactor backend option.

Fixes: https://tracker.ceph.com/issues/69057
Signed-off-by: Mohit Agrawal <moagrawa@redhat.com>

Merge pull request #67044 from Hezko/nvme-submodule-to-1-6-5

mgr/dashboard: bump nvmeof submodule to 1.6.5