cryptsetup is solely used by ceph-volume, so it should be a runtime
dependency of ceph-volume, not ceph-base.
and remove it fomr build-dependency list, as we don't use this tool when
building ceph.
Signed-off-by: Kefu Chai <kchai@redhat.com>
ceph-volume is a tool implemented in pure python, so it would be better
to make it a architecture independent package for better
maintainability.
Signed-off-by: Kefu Chai <kchai@redhat.com>
ceph-volume is a tool implemented in pure python, so it would be better
to make it a architecture independent package for better
maintainability.
in this change
* ceph-volume is extracted out into a separated package
* ceph-volume depends on ceph-osd, as it deploys it and relies on
an already-installed ceph-osd in the system.
* ceph-osd recommends ceph-volume. as ceph-osd can be used as a
standalone package. but ceph-volume enhances it. also, to ensure
the existing users to get ceph-volume installed along with
ceph-osd.
Signed-off-by: Kefu Chai <kchai@redhat.com>
we don't run parted when building ceph, instead, it is a runtime
dependency of ceph-volume, which is packaged in ceph-osd.
so in this change
* parted is removed from build dependency list
* parted is added as a runtime depedency of ceph-osd.
Signed-off-by: Kefu Chai <kchai@redhat.com>
crimson/seastore: add lba tree metrics for lba tree profiling
Reviewed-by: Samuel Just <sjust@redhat.com>
Reviewed-by: Yingxin Cheng <yingxin.cheng@intel.com>
Setting nlextent's to the addr of the original extent violates invariants
in the intrusive_set we're using for ExtentIndex and generally breaks the
assumption in cache that paddrs are unique until reuse. This means we'll
do an extra update_mapping for now -- we can refactor SegmentCleaner to
avoid it later.
Signed-off-by: Samuel Just <sjust@redhat.com>
With this change, Transaction::write_set is the union of mutually exclusive
lists
- delayed_alloc_list, inline_block_list, and mutated_block_list prior to
calling ExtentPlacementManager::delayed_alloc_or_ool_write
and
- inline_block_list, ool_block_list, and mutated_block_list after
calling ExtentPlacementManager::delayed_alloc_or_ool_write
Signed-off-by: Samuel Just <sjust@redhat.com>
delayed_set isn't necessary. Delayed extents have distinct paddr segments
already, so it's simpler to have a single index for all pending extents.
Signed-off-by: Samuel Just <sjust@redhat.com>
This is intended to enchance readability and sligthly
generalize the previous, `_carriage`-aware optimization.
Signed-off-by: Radoslaw Zarzynski <rzarzyns@redhat.com>
It could be that `_carriage` points on empty bptr which is
also stored in the `_buffers` container -- this happens
when we'are trying to preserve the appendable space to not
waste memory and avoid new allocation.
Before the change this approach was imposing unncessary
rebuilding on `c_str()`. Now we take into consideration
the special case when a bufferlist has the extra, empty
bptr at its end.
Fixes: https://tracker.ceph.com/issues/51725
Signed-off-by: Radoslaw Zarzynski <rzarzyns@redhat.com>
The current check only allows to request an OSD id that exists but
marked as 'destroyed'.
With this small fix, we can now use `--osd-id` with an id that doesn't
exist at all.
Fixes: https://tracker.ceph.com/issues/50880
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
osd/scrub: destruct the scrubber shortly before the PG is destructed
Reviewed-by: Kefu Chai <kchai@redhat.com>
Reviewed-by: Samuel Just <sjust@redhat.com>-
Reviewed-by: Neha Ojha <nojha@redhat.com>
When retiring, m_blocks_to_log_entries doesn't remove
corresponding write_entry (should be `*it` not `entry`)
that will be retired. It leads to read error. And
there should also consider discard entries.
Fixes: https://tracker.ceph.com/issues/52579
Signed-off-by: Feng Hualong <hualong.feng@intel.com>
mgr/dashboard: use -f for npm ci to skip fsevents error
Reviewed-by: a16bitsysop <NOT@FOUND>
Reviewed-by: Ernesto Puerta <epuertat@redhat.com>
Reviewed-by: Pere Diaz Bou <pdiazbou@redhat.com>
