cc @pritha-srivastava since you worked on the roles stuff in #43597. i ended up refactoring a lot of that because i have future plans for it. in my work on iam accounts, i'll be making it so that roles can be owned by an account. so in addition to a global listing of roles, you'll also be able to list the roles of a single account for the iam ListRoles api. #54563 adds the cls_user interfaces to track those lists in omap, with similar 'path prefix' filtering to what you wrote for get_roles(). whenever we write/delete/replicate a role in an account, we'll need to update the list in cls_user accordingly

with respect to the creation of mdlog entries, RGWSI_MetaBackend_SObj had taken care of that with its pre_modify()/post_modify(). these mdlog entries must be written manually now, using the new RGWSI_MDLog::complete_entry(). this is only called after a successful write, instead of logging 'prepare' and 'complete' entries separately (metadata sync has always ignored 'prepare' entries)

cc @smanjara on this part

cc @pritha-srivastava since you worked on the roles stuff in #43597. i ended up refactoring a lot of that because i have future plans for it. in my work on iam accounts, i'll be making it so that roles can be owned by an account. so in addition to a global listing of roles, you'll also be able to list the roles of a single account for the iam ListRoles api. #54563 adds the cls_user interfaces to track those lists in omap, with similar 'path prefix' filtering to what you wrote for get_roles(). whenever we write/delete/replicate a role in an account, we'll need to update the list in cls_user accordingly

@cbodley: I reviewed the roles part - all looks fine to me. Just to let you know that currently also listing of roles is for a tenant.

This pull request can no longer be automatically merged: a rebase is needed and changes have to be manually resolved

planning to rebase this once #55152 merges

planning to rebase this once #55152 merges

that did merge, but i'd still like to wait for the squid release so we don't complicate any urgent backports

@cbodley i'm guessing that this will be ready before my OIDC metadata handler changes complete. I probably should rebase on top of this pr?

@cbodley i'm guessing that this will be ready before my OIDC metadata handler changes complete. I probably should rebase on top of this pr?

i don't think this PR touches oidc stuff, so that's probably not necessary. just don't use any 'metadata backend' stuff to implement that

This pull request has been automatically marked as stale because it has not had any activity for 60 days. It will be closed if no further activity occurs for another 30 days.
If you are a maintainer or core committer, please follow-up on this pull request to identify what steps should be taken by the author to move this proposed change forward.
If you are the author of this pull request, thank you for your proposed contribution. If you believe this change is still appropriate, please ensure that any feedback has been addressed and ask for a code review.

This pull request has been automatically closed because there has been no activity for 90 days. Please feel free to reopen this pull request (or open a new one) if the proposed change is still appropriate. Thank you for your contribution!

failed qa in https://pulpito.ceph.com/cbodley-2024-09-07_03:10:16-rgw-wip-rgw-meta-handlers-distro-default-smithi/ with rerun https://pulpito.ceph.com/cbodley-2024-09-07_13:15:19-rgw-wip-rgw-meta-handlers-distro-default-smithi/

multisite job showing crashes in radosgw-admin bilog autotrim:

./src/rgw/driver/rados/rgw_trim_bilog.cc: In function 'virtual int AsyncMetadataList::_send_request(const DoutPrefixProvider*)' thread 7f0aaffe7640 time 2024-09-07T04:04:04.880902+0000
./src/rgw/driver/rados/rgw_trim_bilog.cc: 937: FAILED ceph_assert(keys.size() == 1)

that causes failures for:

FAIL: rgw_multi.tests.test_bucket_index_log_trim
FAIL: rgw_multi.tests.test_replication_status

a notification test is also failing, unrelated to bilog trim:

FAIL: rgw_multi.tests.test_topic_notification_sync

passed qa in https://pulpito.ceph.com/cbodley-2024-09-12_15:39:52-rgw-wip-rgw-meta-handlers-distro-default-smithi/ with rerun https://pulpito.ceph.com/cbodley-2024-09-12_19:58:28-rgw-wip-rgw-meta-handlers-distro-default-smithi/

rebased/squashed

https://jenkins.ceph.com/job/ceph-api/81524/

Command failed with status 250: '../src/vstart.sh -n --nolockdep'

jenkins test api

while testing #53799 after a rebase, I see a crash while removing bucket instance:

2024-09-18T03:26:45.143-0400 7f8474a506c0 20 rgw async rados processor: do_read_bucket_instance_info, bucket instance not found (key=bucket1:41b6e36e-480d-4b64-baec-f43c3a4b73ff.4298.1)

2024-09-18T03:26:45.172-0400 7f8474a506c0 -1 *** Caught signal (Segmentation fault) **
in thread 7f8474a506c0 thread_name:rados_async

ceph version 19.3.0-5006-ge19dc9e23a4 (https://github.com/ceph/ceph/commit/e19dc9e23a4853e4e0e60badddce5448c1c27cf8) squid (dev)
1: /root/ceph/build/bin/radosgw(+0x2172cbb) [0x55f15bdb1cbb]
2: /lib64/libc.so.6(+0x3dbb0) [0x7f858d05fbb0]
3: pthread_mutex_lock()
4: (std::mutex::lock()+0x9) [0x7f858eaa0659]
5: (std::unique_lockstd::mutex::lock()+0x17) [0x7f858eaa08e7]
6: (ceph::logging::Log::submit_entry(ceph::logging::Entry&&)+0x31) [0x7f858ed2bfa1]
7: (RGWMetadataManager::remove(std::__cxx11::basic_string<char, std::char_traits, std::allocator >&, optional_yield, DoutPrefixProvider const*)+0x1df) [0x55f15ba7fc6b]
8: (RGWAsyncMetaRemoveEntry::_send_request(DoutPrefixProvider const*)+0x48) [0x55f15b94bf40]
9: (RGWAsyncRadosRequest::send_request(DoutPrefixProvider const*)+0x1d) [0x55f15baffb35]
10: (RGWAsyncRadosProcessor::handle_request(DoutPrefixProvider const*, RGWAsyncRadosRequest*)+0xc) [0x55f
15baf5c5a]
11: (RGWAsyncRadosProcessor::RGWWQ::_process(RGWAsyncRadosRequest*, ThreadPool::TPHandle&)+0x13) [0x55f15
bafe2a3]
12: (ThreadPool::WorkQueue::_void_process(void*, ThreadPool::TPHandle&)+0xa) [0x55f
15b933b1c]

this was probably never exercised because we didn't remove bucket instance when multisite sync is involved.

while testing #53799 after a rebase, I see a crash while removing bucket instance:

2024-09-18T03:26:45.143-0400 7f8474a506c0 20 rgw async rados processor: do_read_bucket_instance_info, bucket instance not found (key=bucket1:41b6e36e-480d-4b64-baec-f43c3a4b73ff.4298.1)

2024-09-18T03:26:45.172-0400 7f8474a506c0 -1 *** Caught signal (Segmentation fault) ** in thread 7f8474a506c0 thread_name:rados_async

ceph version 19.3.0-5006-ge19dc9e23a4 (https://github.com/ceph/ceph/commit/e19dc9e23a4853e4e0e60badddce5448c1c27cf8) squid (dev)
1: /root/ceph/build/bin/radosgw(+0x2172cbb) [0x55f15bdb1cbb]
2: /lib64/libc.so.6(+0x3dbb0) [0x7f858d05fbb0]
3: pthread_mutex_lock()
4: (std::mutex::lock()+0x9) [0x7f858eaa0659]
5: (std::unique_lockstd::mutex::lock()+0x17) [0x7f858eaa08e7]
6: (ceph::logging::Log::submit_entry(ceph::logging::Entry&&)+0x31) [0x7f858ed2bfa1]
7: (RGWMetadataManager::remove(std::__cxx11::basic_string<char, std::char_traits, std::allocator >&, optional_yield, DoutPrefixProvider const*)+0x1df) [0x55f15ba7fc6b]
8: (RGWAsyncMetaRemoveEntry::_send_request(DoutPrefixProvider const*)+0x48) [0x55f15b94bf40]
9: (RGWAsyncRadosRequest::send_request(DoutPrefixProvider const*)+0x1d) [0x55f15baffb35]
10: (RGWAsyncRadosProcessor::handle_request(DoutPrefixProvider const*, RGWAsyncRadosRequest*)+0xc) [0x55f
15baf5c5a]
11: (RGWAsyncRadosProcessor::RGWWQ::_process(RGWAsyncRadosRequest*, ThreadPool::TPHandle&)+0x13) [0x55f15
bafe2a3]
12: (ThreadPool::WorkQueue::_void_process(void*, ThreadPool::TPHandle&)+0xa) [0x55f
15b933b1c]

this was probably never exercised because we didn't remove bucket instance when multisite sync is involved.

ignore this. i was trying to log something in the RGWMetadataManager without initializing cct.