For replica, filelock in LOCK_LOCK state doesn't allow Fc cap. So
filelock in LOCK_SYNC_LOCK/LOCK_EXCL_LOCK state shouldn't allow Fc
cap either.
Signed-off-by: Yan, Zheng <zheng.z.yan@intel.com>
When inode is freezing or frozen, we defer processing MClientCaps
messages and cap release embedded in requests. The same deferral
logical should also cover MClientCapRelease messages.
After sending cache rejoin message, replica need notify auth MDS when
cap_wanted changes. But it can send MInodeFileCaps message only after
receiving auth MDS' rejoin ack. Locker::request_inode_file_caps() has
correct wait logical, but it skips sending MInodeFileCaps message if
the auth MDS is still in rejoin state.
The fix is defer sending MInodeFileCaps message until the auth MDS
is active. It makes the function's wait logical less tricky.
Signed-off-by: Yan, Zheng <zheng.z.yan@intel.com>
CInode:mds_caps_wanted is used to keep track of caps wanted by non-auth
MDS. The auth MDS checks it when choosing locks' states.
Signed-off-by: Yan, Zheng <zheng.z.yan@intel.com>
when failure of peer is detected, MDCache::handle_mds_failure()
checks if there are requests waiting for slave replies from the
failed peer, and adds them to the "wait for active peer" list.
The "retry request" logical only covers slave requests sent before
MDCache::handle_mds_failure() is called. If a slave request was
sent while peer isn't up, we wait for its reply forever.
Signed-off-by: Yan, Zheng <zheng.z.yan@intel.com>
I previously added code to handle a corner case of cache rejoin:
entire subtree, together with the inode subtree root belongs to,
were trimmed between sending cache rejoin and receiving rejoin ack.
In this case, we should send cache expire message to the subtree's
auth MDS. But the code is complete broken, remove it temporarily.
Signed-off-by: Yan, Zheng <zheng.z.yan@intel.com>
Current code uses import state to detect obsolete import discover/prep
message. it does not work for the case: cancel a subtree import, import
the same subtree again, the discover/prep message for the first import
get dispatched.
Signed-off-by: Yan, Zheng <zheng.z.yan@intel.com>
For unlink/rename request, the target dentry's linkage may change
before all locks are acquired. So we need check if the existing stray
dentry is valid.
Signed-off-by: Yan, Zheng <zheng.z.yan@intel.com>
MDS may crash after journalling a slave commit, but before sending
commit ack to the master. Later when the MDS restarts, it will not
send commit ack to the master. So the master waits for the commit
ack forever. The fix is remove failed MDS from requests' uncommitted
slave list. When failed MDS recovers, its resolve message will tell
the master which slave requests are not committed. The master will
re-add the recovering MDS to requests' uncommitted slave list if
necessary.
Signed-off-by: Yan, Zheng <zheng.z.yan@intel.com>
We may add new waiter while the master is committing. so we should
take the waiters and wake up them when the master is committed.
Signed-off-by: Yan, Zheng <zheng.z.yan@intel.com>
We only journal the finish of exporting subtree, so we shouldn't
consider export bounds as subtree root.
Signed-off-by: Yan, Zheng <zheng.z.yan@intel.com>
Fixes: #5152
When iterating through usage entries, and when user id was
provided, we started at the user's first entry and not from
the entry indexed by the request start time.
This commit fixes the issue.
Backport: bobtail
Signed-off-by: Yehuda Sadeh <yehuda@inktank.com>
Reviewed-by: Greg Farnum <greg@inktank.com>
fixed bug in init script, the "df" should be run on remote host by do_cmd,
and use $host instead of "hostname -s"
Signed-off-by: Xiaoxi Chen <xiaoxi.chen@intel.com>
(cherry picked from commit 1dd99f0fc9)
Conflicts:
src/init-ceph.in
Fixes valgrind warning:
==14803== Use of uninitialised value of size 8
==14803== at 0x12E7614: sctp_crc32c_sb8_64_bit (sctp_crc32.c:567)
==14803== by 0x12E76F8: update_crc32 (sctp_crc32.c:609)
==14803== by 0x12E7720: ceph_crc32c_le (sctp_crc32.c:733)
==14803== by 0x105085F: ceph::buffer::list::crc32c(unsigned int) (buffer.h:427)
==14803== by 0x115D7B2: Message::calc_front_crc() (Message.h:441)
==14803== by 0x1159BB0: Message::encode(unsigned long, bool) (Message.cc:170)
==14803== by 0x1323934: Pipe::writer() (Pipe.cc:1524)
==14803== by 0x13293D9: Pipe::Writer::entry() (Pipe.h:59)
==14803== by 0x120A398: Thread::_entry_func(void*) (Thread.cc:41)
==14803== by 0x503BE99: start_thread (pthread_create.c:308)
==14803== by 0x6C6E4BC: clone (clone.S:112)
Backport: cuttlefish
Signed-off-by: Sage Weil <sage@inktank.com>
We need to do df on the remote host, not locally.
Simlarly, the ceph command uses the osd key, which exists remotely; run it there.
Signed-off-by: Sage Weil <sage@inktank.com>
We would need to do hostname -s on the remote node, not the local one.
But we already have $host; use it!
Reported-by: Xiaoxi Chen <xiaoxi.chen@intel.com>
Signed-off-by: Sage Weil <sage@inktank.com>
New pools won't be full. mon->pgmon()->pg_map.pg_pool_sum[poolid] will
implicitly create an entry for poolid causing register_new_pgs() to assume that
the newly created pgs in the new pool are in fact a result of a split
preventing MOSDPGCreate messages from being sent out.
Fixes: #4813
Backport: cuttlefish
Signed-off-by: Samuel Just <sam.just@inktank.com>
Reviewed-by: Sage Weil <sage@inktank.com>
We don't need it after all. If we are in the middle of some proposal,
then we guarantee that said proposal is likely to be retried. If we
haven't yet proposed, then it's forever more likely that a client will
eventually retry the message that triggered this proposal.
Basically, this mechanism attempted at fixing a non-problem, and was in
fact triggering some unforeseen issues that would have required increasing
the code complexity for no good reason.
Fixes: #5102
Signed-off-by: Joao Eduardo Luis <joao.luis@inktank.com>
By finishing these Contexts, we make sure the Contexts they enclose (to be
called once the proposal goes through) will behave as their were initially
planned: for instance, a C_Command() may retry the command if a -EAGAIN
is passed to 'finish_contexts', while a C_Trimmed() will simply set
'going_to_trim' to false.
This aims at fixing at least a bug in which Paxos will stop trimming if an
election is triggered while a trim is queued but not yet finished. Such
happens because it is the C_Trimmed() context that is responsible for
resetting 'going_to_trim' back to false. By clearing all the contexts on
the proposal list instead of finishing them, we stay forever unable to
trim Paxos again as 'going_to_trim' will stay True till the end of time as
we know it.
Fixes: #4895
Signed-off-by: Joao Eduardo Luis <joao.luis@inktank.com>
We were always disabling compression.
Fixes: #5131
Reported-by: Sylvain Munaut <s.munaut@whatever-company.com>
Signed-off-by: Sage Weil <sage@inktank.com>
- prepend $local to the $allconf list at the top
- remove $local special case for all case
- fix the type prefix checks to explicitly check for prefixes
Fugly bash, but works!
Backport: cuttlefish, bobtail
Signed-off-by: Sage Weil <sage@inktank.com>
Reviewed-by: Dan Mick <dan.mick@inktank.com>
Wheezy's udev (175-7.2) has broken rules for the /dev/disk/by-partuuid/
symlinks that ceph-disk relies on. Install parallel rules that work. On
new udev, this is harmless; old older udev, this will make life better.
Fixes: #4865
Backport: cuttlefish
Signed-off-by: Sage Weil <sage@inktank.com>
If we have not pg_map updates, the delta doesn't update, and can get stuck
with the velocity right before activity stopped. This is confusing, and
can cause incorrect health warnings about in-progress recovery.
To fix this, zero the delta if there is no activity for
'mon delta reset interval' seconds.
Fixes: #5077
Signed-off-by: Sage Weil <sage@inktank.com>
FileStoreDiff umounts the stores in its destructor.
Also, DeterministicOpSequence handles deletes its passed
object store.
Fixes: #5076
Signed-off-by: Samuel Just <sam.just@inktank.com>
Reviewed-by: David Zafman <david.zafman@inktank.com>
The finish() method for Contexts calls delete this.
This reverts commit 36028916c4.
Fixes: #5075
Signed-off-by: Samuel Just <sam.just@inktank.com>
Reviewed-by: David Zafman <david.zafman@inktank.com>
...indirectly via ceph-common. We get bad behavior when they diverge, I
think because of libcommon.la being linked both statically and dynamically.
Fixes: #4997
Backport: cuttlefish, bobtail
Signed-off-by: Sage Weil <sage@inktank.com>
Reviewed-by: Gary Lowell <gary.lowell@inktank.com>
In committed_thru, we use write_pos to reset the header.start value in cases
where seq is past the end of our journalq. It is therefore important that the
journalq be updated atomically with write_pos (that is, under the write_lock).
The call to align_bl() is moved into do_write in order to ensure that write_pos
is adjusted correctly prior to write_bl().
Also, we adjust pos at the end of write_bl() such that pos \in [get_top(),
header.max_size) after write_bl().
Fixes: #5020
Signed-off-by: Samuel Just <sam.just@inktank.com>
Reviewed-by: Sage Weil <sage@inktank.com>
