PeeringState was the only remaining caller -- both the interface in
PGBackend and the pass-through in PrimaryLogPG were actually unused.
Further, the WaitLocalRecoveryReserved user does not appear to actually
require the fresh-off-of-the-wire osdmap available from OSDService.
As such, moved the logic into OSDMap itself and simply used the
PG local osdmap.
Note, the above is a change in behavior that probably could use a second
opinion.
Samuel Just [Fri, 29 Mar 2019 18:06:26 +0000 (11:06 -0700)]
osd/: move Active purged_snaps handling back to PG
Note: this patch only sets dirty_info/dirty_big_info when
info.purged_snaps is actually changed. I *think* that's
right, but that part deserves particular attention during
review.
osd/: move io reservation machinery into PeeringState
This patch recasts the reservation and backoff interfaces
on PeeringListener in terms of events to queue rather than
explicit callbacks (PG handles implementing via callbacks).
sjust@redhat.com [Fri, 22 Mar 2019 23:19:18 +0000 (16:19 -0700)]
osd/: move most peering state to PeeringState
This patch moves the 40 something peering state variables over into
PeeringState while leaving references to them in PG. The following
patches will move over the users until all users are in PeeringState.
Then, the PG references will be removed. A subsequent patch will also
move the recovery_state member to be the last initialize and first
destructed.
Samuel Just [Thu, 21 Mar 2019 23:25:17 +0000 (16:25 -0700)]
PGStateUtils: remove PG*, move into PGStateHistory into PeeringState
I don't think there's any need to worry about the pg locking from
PGStateHistory. NamedState::begin/exit and dump_pgstate_history are the
only users, so the only calls should be under the peering event handler,
the asok command, or the PG constructor and destructor. The first two
already have the lock. The last should be safe as well as long as
the state machine states are constructed and destructed after and
before the PGStateHistory instance. As such, this patch removes most
of that state leaving the epoch generation as an interface
implemented by PG.
The snap trimming state machine was already excluded, so that this
patch leaves it disabled.
Samuel Just [Fri, 12 Apr 2019 18:08:54 +0000 (11:08 -0700)]
osd/: mechanically rename RecoveryMachine/State/Ctx to Peering*
I'm going to extract this logic and reuse it in crimson. Recovery* has
always been a confusing name as it implements neither log-based recovery
nor backfill. Rather, it's mainly the buisiness logic for agreeing on
an authoritative log and some ancillary things such as scrub/backfill
reservation.
$ for i in $(git grep -l 'RecoveryMachine'); do sed -i 's/RecoveryMachine/PeeringMachine/g' $i; done
$ for i in $(git grep -l 'RecoveryState'); do sed -i 's/RecoveryState/PeeringState/g' $i; done
$ for i in $(git grep -l 'RecoveryCtx'); do sed -i 's/RecoveryCtx/PeeringCtx/g' $i; done
Sage Weil [Mon, 29 Apr 2019 16:04:15 +0000 (11:04 -0500)]
Merge PR #27804 into master
* refs/pull/27804/head:
common, osd: remove erroneous 'section' in dump functions
dencoder: include some missed types
common: make utime_t dencoder-compatible
common: make uuid_d dencoder-compatible
osd: make watch_item_t dencoder-compatible
crimson/mon: hold rotating_keyring using a pointer
we could have kept rotating_keyring as a plain member variable, but
`AuthClientHandler` keeps a weak reference to it, we are not able to
update it to point it to the new rotating_keyring in the newly created
active_con created from the connected pending connection -- the move
ctor of ceph::mon::Connection does create a new RotatingKeyring. so this
has two consequences:
- the raw pointer held by AuthClientHandler is not valid anymore, after
the pending connection is destroyed when it is promoted to active_con.
so we are writing to freed memory when renewing the rotating keyring.
- we won't have access to the updated keyring.
if we use a std::vector<unique_ptr<Connection>> for pending_conns, this
change won't be needed. we could make this change in a future change
though.
auth,mon,crimson: pass KeyStore by const reference
AuthAuthorizeHandler::verify_authorizer() neither changes the keystore,
nor expects a nullptr. so we should pass the keystore by const reference
for better readability
crimson-common and ceph-common have subtle difference which is
conditionalized by WITH_SEASTAR macro. and the different implementations
share the same name, so we should not link crimson executables against ceph-common .
crimson messenger's tests does not use ConfigProxy at the time of writing,
and ldout() cannot tell if a log message should be written to the
logfile or not without querying config subsystem, so, to avoid involving
config system in crimson messenger's tests, the log line is disabled if
WITH_SEASTAR, we will kill this workaround once we ditch crimson/net/Config.h.
auth/cephx: do not require a cct for accessing config
crimson/msgr does not depend on ConfigProxy at the time of writing. we
want to decouple it from config subsystem in hope to have less cross
subsystem dependencies and speed up the compiling speed.
but, to introduce cephx to msgr, we need to enable msgr to read from
ConfigProxy at runtime. so, to avoid keeping a CephContext instance
simply for accessing the ConfigProxy reference in it, it would be more
efficient to define a free function `conf()` which hides the difference
between crimson and classic user of auth/cephx.
CephxAuthorizeHandler::verify_authorizer() uses it for querying config
and for a RNG, so we have to pass it a CephContext which has `_conf` and
`random()`.
This flag is used for compatibility with pre-nautilus OSDs, which do not
send authorizers on the OSD heartbeat connections. However, because the
AuthServer is implemented by MonClient, which is shared across all
OSD messengers, we can't set this to false for the OSD without disabling
all auth. Instead, make it a Messenger property, and set it only on the
heartbeat server messengers.
Sage Weil [Sun, 28 Apr 2019 13:43:28 +0000 (08:43 -0500)]
msg: set_require_authorizer on messenger, not dispatcher
This flag is used for compatibility with pre-nautilus OSDs, which do not
send authorizers on the OSD heartbeat connections. However, because the
AuthServer is implemented by MonClient, which is shared across all
OSD messengers, we can't set this to false for the OSD without disabling
all auth. Instead, make it a Messenger property, and set it only on the
heartbeat server messengers.
Jason Dillaman [Fri, 19 Apr 2019 18:46:02 +0000 (14:46 -0400)]
librbd: removed 'ImageCtx::md_lock'
This lock used to protect the IO pathway to prevent writes but
that is now handled by the io::ImageRequestWQ. Additional
historical uses have been temporarily moved to the
'ImageCtx::image_lock'
Signed-off-by: Jason Dillaman <dillaman@redhat.com>
crimson/net: prefer <fmt/chrono.h> over <fmt/time.h>
in latest libfmt, <fmt/time.h> is deprecated.
to silence warnings like
/home/kchai/ceph/src/fmt/include/fmt/time.h:13:2: warning: #warning
fmt/time.h is deprecated, use fmt/chrono.h instead [-Wcpp]
#warning fmt/time.h is deprecated, use fmt/chrono.h instead
^~~~~~~
In file included from
/home/kchai/ceph/src/seastar/include/seastar/core/reactor.hh:72:0,
from
/home/kchai/ceph/src/seastar/include/seastar/core/sharded.hh:24,
from /home/kchai/ceph/src/crimson/net/Fwd.h:18,
from /home/kchai/ceph/src/crimson/net/Protocol.h:9,
from /home/kchai/ceph/src/crimson/net/ProtocolV2.h:6,
from /home/kchai/ceph/src/crimson/net/ProtocolV2.cc:4:
projects / ceph.git / log