PGRecovery::on_global_recover destroys the map entry without waiting for
the future returned from
seastar::future<> wait_for_recovered(BlockingEvent::TriggerI& trigger) {
This commit changes WaitForObjectRecovery to be refcounted and retains a
reference until the future resolves.
Fixes: https://tracker.ceph.com/issues/55565
Signed-off-by: Samuel Just <sjust@redhat.com>
The rgw_admin_curl.sh script will allow end-user/developers to
access RGW admin APIs through curl command.
Signed-off-by: Prashant D <pdhange@redhat.com>
DBStore will use the below options when creating .db file for rgw server
to use.
'dbstore_db_dir':
- path for the directory to store db backend store files
- default: /var/run/ceph
'dbstore_db_name_prefix':
- prefix to be used for the files created by db backend store
- defauult: dbstore
For eg., by default the full path of the file shall be
eg., /var/run/ceph/dbstore-default_ns.db
Signed-off-by: Soumya Koduri <skoduri@redhat.com>
By the introduction of range blocklist, the 'blocklist ls' command outputs
two lists. It's also straightforward to get the blocklisted clients directly
from 'osd dump' to avoid regression.
Fixes: https://tracker.ceph.com/issues/55516
Signed-off-by: Jos Collin <jcollin@redhat.com>
We're going to unify responsibility for routing requests to requests
to start_pg_operation from the different osd operation types. Later,
this will move into a new type responsible for dealing with the pg->core
mapping.
Signed-off-by: Samuel Just <sjust@redhat.com>
Really, all pg routable events (peering events, client operations, etc)
should have the same sequence of pre-pg blockers.
Signed-off-by: Samuel Just <sjust@redhat.com>
stop.sh removes the configured asok_dir, so defaulting this to out/
results in logs being removed. Let's instead just create an
asok/ directory.
Fixes: https://tracker.ceph.com/issues/55563
Signed-off-by: Samuel Just <sjust@redhat.com>
A bug in AWS documentation caused this to be named "NoSuchTagSetError",
when the proper name (now fixed in the docs) is "NoSuchTagSet". Fix
this so we match AWS.
Fixes: https://tracker.ceph.com/issues/55460
Author: Peter Gerber <peter@arbitrary.ch>
Signed-off-by: Daniel Gryniewicz <dang@redhat.com>
This methods fails to collect return value from
FuseMount._run_mount_cmd() and return it. This leads to a bug for tests
that expect mount command to fail when executed with vstart_runner.py.
Fixes: https://tracker.ceph.com/issues/55553
Signed-off-by: Rishabh Dave <ridave@redhat.com>
And therefore get rid of methods duplicated in LocalRemote and add a
call to empty constructor of RemoteShell in LocalRemote.__init__().
Signed-off-by: Rishabh Dave <ridave@redhat.com>
The optimization isn't the goal (but rather a side effect). The
reason behind this change is to let pipeline stages define exit
barriers incorporating `OpT`-dependant `Triggers` to track blocking
that happens on the barriers.
Signed-off-by: Radosław Zarzyński <rzarzyns@redhat.com>
