Sage Weil [Fri, 31 May 2013 23:39:37 +0000 (16:39 -0700)]
mon/Paxos: go active *after* refreshing
The update_from_paxos() methods occasionally like to trigger new activity.
As long as they check is_readable() and is_writeable(), they will defer
until we go active and that activity will happen in the normal callbacks.
This fixes the problem where we active but is_writeable() is still false,
triggered by PGMonitor::check_osd_map().
Sage Weil [Sun, 2 Jun 2013 23:57:11 +0000 (16:57 -0700)]
mon/Paxos: do paxos refresh in finish_proposal; and refactor
Do the paxos refresh inside finish_proposal, ordered *after* the leader
assertion so that MonmapMonitor::update_from_paxos() calling bootstrap()
does not kill us.
Also, remove unnecessary finish_queued_proposal() and move the logic inline
where the bad leader assertion is obvious.
Sage Weil [Sun, 2 Jun 2013 23:14:01 +0000 (16:14 -0700)]
mon: explicitly refresh_from_paxos() when leveldb state changes
Instead of opportunistically calling each service's update_from_paxos(),
instead explicitly refresh all in-memory state whenever we know the
paxos state may have changed. This is simpler and less fragile.
Sage Weil [Thu, 9 May 2013 16:44:20 +0000 (09:44 -0700)]
osd: init test_ops_hook
CID 1019628 (#1 of 1): Uninitialized pointer field (UNINIT_CTOR)
2. uninit_member: Non-static class member "test_ops_hook" is not initialized in this constructor nor in any functions that it calls.
Sage Weil [Thu, 9 May 2013 16:45:51 +0000 (09:45 -0700)]
osd: initialize OSDService::next_notif_id
CID 1019627 (#1 of 1): Uninitialized scalar field (UNINIT_CTOR)
2. uninit_member: Non-static class member "next_notif_id" is not initialized in this constructor nor in any functions that it calls.
Sage Weil [Wed, 22 May 2013 21:29:37 +0000 (14:29 -0700)]
messages/MOSDMarkMeDown: fix uninit field
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)
Sage Weil [Tue, 18 Jun 2013 03:32:15 +0000 (20:32 -0700)]
common/Preforker: fix warning
common/Preforker.h: In member function ‘int Preforker::signal_exit(int)’:
warning: common/Preforker.h:82:45: ignoring return value of ‘ssize_t safe_write(int, const void*, size_t)’, declared with attribute warn_unused_result [-Wunused-result]
This is harder than it should be to fix. :(
http://stackoverflow.com/questions/3614691/casting-to-void-doesnt-remove-warn-unused-result-error
Whatever, I guess we can do something useful with this return value.
Signed-off-by: Sage Weil <sage@inktank.com> Reviewed-by: David Zafman <david.zafman@inktank.com>
(cherry picked from commit ce7b5ea7d5c30be32e4448ab0e7e6bb6147af548)
mon: Monitor: make sure we backup a monmap during sync start
First of all, we must find a monmap to backup. The newest version.
Secondly, we must make sure we back it up before clearing the store.
Finally, we must make sure that we don't remove said backup while
clearing the store; otherwise, we would be out of a backup monmap if the
sync happened to fail (and if the monitor happened to be killed before a
new sync had finished).
This patch makes sure these conditions are met.
Fixes: #5256 (partially)
Backport: cuttlefish
Signed-off-by: Joao Eduardo Luis <joao.luis@inktank.com> Reviewed-by: Sage Weil <sage@inktank.com>
(cherry picked from commit 5e6dc4ea21b452e34599678792cd36ce1ba3edb3)
mon: Monitor: obtain latest monmap on sync store init
Always use the highest version amongst all the typically available
monmaps: whatever we have in memory, whatever we have under the
MonmapMonitor's store, and whatever we have backed up from a previous
sync. This ensures we always use the newest version we came across
with.
Signed-off-by: Joao Eduardo Luis <joao.luis@inktank.com> Reviewed-by: Sage Weil <sage@inktank.com>
(cherry picked from commit 6284fdce794b73adcc757fee910e975b6b4bd054)
mon: Monitor: don't remove 'mon_sync' when clearing the store during abort
Otherwise, we will end up losing the monmap we backed up when we started
the sync, and the monitor may be unable to start if it is killed or
crashes in-between the sync abort and finishing a new sync.
Fixes: #5256 (partially)
Backport: cuttlefish
Signed-off-by: Joao Eduardo Luis <joao.luis@inktank.com> Reviewed-by: Sage Weil <sage@inktank.com>
(cherry picked from commit af5a9861d7c6b4527b0d2312d0efa792910bafd9)
Sage Weil [Wed, 19 Jun 2013 04:31:23 +0000 (21:31 -0700)]
os/FileStore: drop posix_fadvise(...DONTNEED)
On XFS this call is problematic because it directly calls the filemap
writeback without vectoring through xfs. This can break the delicate
ordering of writeback and range zeroing; see #4976 and this thread
Sage Weil [Sun, 16 Jun 2013 03:06:33 +0000 (20:06 -0700)]
ceph-disk: do not stop activate-all on first failure
Keep going even if we hit one activation error. This avoids failing to
start some disks when only one of them won't start (e.g., because it
doesn't belong to the current cluster).
Sage Weil [Fri, 14 Jun 2013 22:01:14 +0000 (15:01 -0700)]
ceph.spec: install/uninstall init script
This was commented out almost years ago in commit 9baf5ef4 but it is not
clear to me that it was correct to do so. In any case, we are not
installing the rc.d links for ceph, which means it does not start up after
a reboot.
Sage Weil [Fri, 14 Jun 2013 20:34:40 +0000 (13:34 -0700)]
ceph-disk: add 'activate-all'
Scan /dev/disk/by-parttypeuuid for ceph OSDs and activate them all. This
is useful when the event didn't trigger on the initial udev event for
some reason.
Yehuda Sadeh [Fri, 14 Jun 2013 21:53:54 +0000 (14:53 -0700)]
rgw: escape prefix correctly when listing objects
Fixes: #5362
When listing objects prefix needs to be escaped correctly (the
same as with the marker). Otherwise listing objects with prefix
that starts with underscore doesn't work.
Backport: bobtail, cuttlefish
Sage Weil [Sat, 15 Jun 2013 15:14:40 +0000 (08:14 -0700)]
common/Preforker: fix broken recursion on exit(3)
If we exit via preforker, call exit(3) and not recursively back into
Preforker::exit(r). Otherwise you get a hang with the child blocked
at:
Thread 1 (Thread 0x7fa08962e7c0 (LWP 5419)):
#0 0x000000309860e0cd in write () from /lib64/libpthread.so.0
#1 0x00000000005cc906 in Preforker::exit(int) ()
#2 0x00000000005c8dfb in main ()
and the parent at
#0 0x000000309860eba7 in waitpid () from /lib64/libpthread.so.0
#1 0x00000000005cc87a in Preforker::parent_wait() ()
#2 0x00000000005c75ae in main ()
Sage Weil [Fri, 14 Jun 2013 04:56:23 +0000 (21:56 -0700)]
ceph-disk: do not use mount --move (or --bind)
The kernel does not let you mount --move when the parent mount is
shared (see, e.g., https://bugzilla.redhat.com/show_bug.cgi?id=917008
for another person this also confused). We can't use --bind either
since that (on RHEL at least) screws up /etc/mtab so that the final
result looks like
Sage Weil [Thu, 13 Jun 2013 22:54:58 +0000 (15:54 -0700)]
ceph-disk: implement 'activate-journal'
Activate an osd via its journal device. udev populates its symlinks and
triggers events in an order that is not related to whether the device is
an osd data partition or a journal. That means that triggering
'ceph-disk activate' can happen before the journal (or journal symlink)
is present and then fail.
Similarly, it may be that they are on different disks that are hotplugged
with the journal second.
This can be wired up to the journal partition type to ensure that osds are
started when the journal appears second.
Sage Weil [Wed, 12 Jun 2013 01:35:01 +0000 (18:35 -0700)]
ceph-disk: call partprobe outside of the prepare lock; drop udevadm settle
After we change the final partition type, sgdisk may or may not trigger a
udev event, depending on how well udev is behaving (it varies between
distros, it seems). The old code would often settle and wait for udev to
activate the device, and then partprobe would uselessly fail because it
was already mounted.
Call partprobe only at the very end, after prepare is done. This ensures
that if partprobe calls udevadm settle (which is sometimes does) we do not
get stuck.
Drop the udevadm settle. I'm not sure what this accomplishes; take it out,
at least until we determine we need it.
It is often useful to prepare but not activate a device, for example when
preparing a bunch of spare disks. This marks a device as 'do not
activate' so that it can be prepared without activating.
Sage Weil [Fri, 14 Jun 2013 18:21:25 +0000 (11:21 -0700)]
upstart: start ceph-all on runlevel [2345]
Starting when only one network interface has started breaks machines with
multiple nics in very problematic ways.
There may be an earlier trigger that we can use for cases where other
services on the local machine depend on ceph, but for now this is better
than the existing behavior.
Sage Weil [Sun, 9 Jun 2013 00:38:07 +0000 (17:38 -0700)]
client: set issue_seq (not seq) in cap release
We regularly have been observing a stall where the MDS is blocked waiting
for a cap revocation (Ls, in our case) and never gets a reply. We finally
tracked down the sequence:
- mds issues cap seq 1 to client
- mds does revocation (seq 2)
- client replies
- much time goes by
- client trims inode from cache, sends release with seq == 2
- mds ignores release because its issue_seq is 1
- mds later tries to revoke other caps
- client discards message because it doesn't have the inode in cache
The problem is simply that we are using seq instead of issue_seq in the
cap release message. Note that the other release call site in
encode_inode_release() is correct. That one is much more commonly
triggered by short tests, as compared to this case where the inode needs to
get pushed out of the client cache.
Sage Weil [Mon, 3 Jun 2013 04:21:51 +0000 (21:21 -0700)]
ceph-fuse: create finisher threads after fork()
The ObjectCacher and MonClient classes both instantiate Finisher
threads. We need to make sure they are created *after* the fork(2)
or else the process will fail to join() them on shutdown, and the
threads will not exist while fuse is doing useful work.
Put CephFuse on the heap and move all this initalization into the child
block, and make sure errors are passed back to the parent.
Fix-proposed-by: Alexandre Marangone <alexandre.maragone@inktank.com> Signed-off-by: Sage Weil <sage@inktank.com>
(cherry picked from commit 4fa5f99a40792341d247e51488c37301da3c4e4f)
Sage Weil [Thu, 6 Jun 2013 23:35:54 +0000 (16:35 -0700)]
osd: do not include logbl in scrub map
This is a potentially use object/file, usually prefixed by a zeroed region
on disk, that is not used by scrub at all. It dates back to f51348dc8bdd5071b7baaf3f0e4d2e0496618f08 (2008) and the original version of
scrub.
This *might* fix #4179. It is not a leak per se, but I observed 1GB
scrub messages going over the write. Maybe the allocations are causing
fragmentation, or the sub_op queues are growing.
Signed-off-by: Sage Weil <sage@inktank.com> Reviewed-by: Samuel Just <sam.just@inktank.com>
(cherry picked from commit 0b036ecddbfd82e651666326d6f16b3c000ade18)
Yehuda Sadeh [Fri, 7 Jun 2013 04:53:00 +0000 (21:53 -0700)]
rgw: handle deep uri resources
In case of deep uri resources (ones created beyond a single level
of hierarchy, e.g. auth/v1.0) we want to create a new empty
handlers for the path if no handlers exists. E.g., for
auth/v1.0 we need to have a handler for 'auth', otherwise
the default S3 handler will be used, which we don't want.
Yehuda Sadeh [Fri, 7 Jun 2013 04:47:21 +0000 (21:47 -0700)]
rgw: fix get_resource_mgr() to correctly identify resource
Fixes: #5262
The original test was not comparing the correct string, ended up
with the effect of just checking the substring of the uri to match
the resource.
Sage Weil [Sat, 1 Jun 2013 04:23:45 +0000 (21:23 -0700)]
mon: fix preforker exit behavior behavior
In 3c5706163b72245768958155d767abf561e6d96d we made exit() not actually
exit so that the leak checking would behave for a non-forking case.
That is only needed for the normal exit case; every other case expects
exit() to actually terminate and not continue execution.
Instead, make a signal_exit() method that signals the parent (if any)
and then lets you return. exit() goes back to it's usual behavior,
fixing the many other calls in main().
Backport: cuttlefish Signed-off-by: Sage Weil <sage@inktank.com> Reviewed-by: Joao Eduardo Luis <joao.luis@inktank.com>
(cherry picked from commit 92d085f7fd6224ffe5b7651c1f83b093f964b5cd)
Josh Durgin [Tue, 4 Jun 2013 20:23:36 +0000 (13:23 -0700)]
rados.py: correct some C types
trunc was getting size_t instead of uint64_t, leading to bad results
in 32-bit environments. Explicitly cast to the desired type
everywhere, so it's clear the correct type is being used.
Sage Weil [Sat, 1 Jun 2013 00:09:19 +0000 (17:09 -0700)]
mon: start lease timer from peon_init()
In the scenario:
- leader wins, peons lose
- leader sees it is too far behind on paxos and bootstraps
- leader tries to sync with someone, waits for a quorum of the others
- peons sit around forever waiting
The problem is that they never time out because paxos never issues a lease,
which is the normal timeout that lets them detect a leader failure.
Avoid this by starting the lease timeout as soon as we lose the election.
The timeout callback just does a bootstrap and does not rely on any other
state.
I see one possible danger here: there may be some "normal" cases where the
leader takes a long time to issue its first lease that we currently
tolerate, but won't with this new check in place. I hope that raising
the lease interval/timeout or reducing the allowed paxos drift will make
that a non-issue. If it is problematic, we will need a separate explicit
"i am alive" from the leader while it is getting ready to issue the lease
to prevent a live-lock.
Backport: cuttlefish, bobtail Signed-off-by: Sage Weil <sage@inktank.com> Reviewed-by: Greg Farnum <greg@inktank.com>
(cherry picked from commit f1ccb2d808453ad7ef619c2faa41a8f6e0077bd9)
Sage Weil [Fri, 31 May 2013 05:52:21 +0000 (22:52 -0700)]
mon: discard messages from disconnected clients
If the client is not connected, discard the message. They will
reconnect and resend anyway, so there is no point in processing it
twice (now and later).
- trim more at a time (by an order of magnitude)
- rename fields to paxos_trim_{min,max}; only trim when there are min items
that are trimmable, and trim at most max items at a time.
- adjust the paxos_service_trim_{min,max} values up by a factor of 2.
Since we are compacting every time we trim, adjusting these up mean less
frequent compactions and less overall work for the monitor.
Sage Weil [Thu, 30 May 2013 21:36:41 +0000 (14:36 -0700)]
mon: make compaction bounds overlap
When we trim items N to M, compact over range (N-1) to M so that the
items in the queue will share bounds and get merged. There is no harm in
compacting over a larger range here when the lower bound is a key that
doesn't exist anyway.
Sage Weil [Wed, 29 May 2013 15:35:44 +0000 (08:35 -0700)]
mon/MonitorDBStore: allow compaction of ranges
Allow a transaction to describe the compaction of a range of keys. Do this
in a backward compatible say, such that older code will interpret the
compaction of a prefix + range as compaction of the entire prefix. This
allows us to avoid introducing any new feature bits.
Sage Weil [Tue, 28 May 2013 23:35:55 +0000 (16:35 -0700)]
os/LevelDBStore: do compact_prefix() work asynchronously
We generally do not want to block while compacting a range of leveldb.
Push the blocking+waiting off to a separate thread. (leveldb will do what
it can to avoid blocking internally; no reason for us to wait explicitly.)
Samuel Just [Mon, 15 Apr 2013 23:33:48 +0000 (16:33 -0700)]
PG: don't write out pg map epoch every handle_activate_map
We don't actually need to write out the pg map epoch on every
activate_map as long as:
a) the osd does not trim past the oldest pg map persisted
b) the pg does update the persisted map epoch from time
to time.
To that end, we now keep a reference to the last map persisted.
The OSD already does not trim past the oldest live OSDMapRef.
Second, handle_activate_map will trim if the difference between
the current map and the last_persisted_map is large enough.
Fixes: #4731 Signed-off-by: Samuel Just <sam.just@inktank.com> Reviewed-by: Greg Farnum <greg@inktank.com>
(cherry picked from commit 2c5a9f0e178843e7ed514708bab137def840ab89)
Conflicts:
src/common/config_opts.h
src/osd/PG.cc
- last_persisted_osdmap_ref gets set in the non-static
PG::write_info
Samuel Just [Tue, 21 May 2013 22:22:56 +0000 (15:22 -0700)]
OSDMonitor: skip new pools in update_pools_status() and get_pools_health()
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>
(cherry picked from commit 0289c445be0269157fa46bbf187c92639a13db46)
Yehuda Sadeh [Thu, 30 May 2013 19:58:11 +0000 (12:58 -0700)]
rgw: only append prefetched data if reading from head
Fixes: #5209
Backport: bobtail, cuttlefish
If the head object wrongfully contains data, but according to the
manifest we don't read from the head, we shouldn't copy the prefetched
data. Also fix the length calculation for that data.
projects / ceph.git / log