Pass in fs_op_seq (last_committed_seq), not the next expected seq, so we
can avoid subtracting and adding 1 in odd places.
Signed-off-by: Sage Weil <sage@newdream.net>
It is perfectly fine to read events that are older than the fs's seq from
the journal; open() will skip them when positioning the read pointer on
open.
Also, this code is nonsensical; it always failed the assertion.
Signed-off-by: Sage Weil <sage@newdream.net>
last_committed_seq is the last seq committed to the fs, not the journal.
Set it when we begin replay with the fs provided value, not from the newest
entry in the journal.
Signed-off-by: Sage Weil <sage@newdream.net>
The 'mon' perfcounter is for the local daemon and is always registered.
The 'cluster' perfcounter is for cluster state, and is only registered
(and thus only shows up via the admin socket) when the current daemon is
part of the cluster quorum.
No actual counters yet.
Signed-off-by: Sage Weil <sage@newdream.net>
This could conceivably cause the reply ordering mismatch seen in bug
#1490. Not sure why we didn't also fix this caller when we fixed that
bug last time :).
Signed-off-by: Sage Weil <sage@newdream.net>
This might happen if, e.g., the file_layout specifies an osd that later
is removed from the cluster entirely. Just ignore it instead of making
upper layers duplicate this check.
Signed-off-by: Sage Weil <sage.weil@dreamhost.com>
This reverts commit 6fbab6da69.
This fails a unit test.
And I change my mind.. I think this is most cleanly handled inside crush, so
we don't duplicate the same check that is generating the error with an different
data structure.
It was not an index, and seems to contain recommendations
for system configuration. I have renamed it to confusing.txt
and will merge it in a future commit.
Signed-off-by: Mark Kampe <mark.kampe@dreamhost.com>
which was a copy of PlanningImplementation.txt
(and not html at all).
restored previous index.rst, which was overwritten with a copy
of PlanninImplementation.txt, but removed all of the recursively
included content from the document.
I will cherry-pick merge the new contents in a subsequent commit.
Signed-off-by: Mark Kampe <mark.kampe@dreamhost.com>
it was accidentally overwritten with a version of the product
had a somewhat different audience/focus and a few sphinx
formatting errors.
I will cherry-pick the corrections in a subsequent commit.
Signed-off-by: Mark Kampe <mark.kampe@dreamhost.com>
We haven't made explicit that the Hadoop Java code is under the Apache
License. Do so (with permission from the other contributors, thanks!).
Signed-off-by: Greg Farnum <gregory.farnum@dreamhost.com>
We need to hold mylock before waiting on the cond or else we get
./common/Cond.h: In function 'int Cond::Wait(Mutex&)', in thread '7f37fe0c8700'
./common/Cond.h: 46: FAILED assert(mutex.is_locked())
ceph version 0.38-2-g73f99a1 (commit:73f99a189f491866da2be88adcfe0bd512282755)
1: (MDLog::_replay_thread()+0x2483) [0x6c4393]
2: (MDLog::ReplayThread::entry()+0xd) [0x4decbd]
3: (()+0x6d8c) [0x7f3803e8fd8c]
4: (clone()+0x6d) [0x7f38028d504d]
ceph version 0.38-2-g73f99a1 (commit:73f99a189f491866da2be88adcfe0bd512282755)
1: (MDLog::_replay_thread()+0x2483) [0x6c4393]
2: (MDLog::ReplayThread::entry()+0xd) [0x4decbd]
3: (()+0x6d8c) [0x7f3803e8fd8c]
4: (clone()+0x6d) [0x7f38028d504d]
*** Caught signal (Aborted) **
in thread 7f37fe0c8700
Signed-off-by: Sage Weil <sage@newdream.net>
This eliminates some flags and avoids annoying cases where the banner is
printed but we don't want to see it.
Signed-off-by: Sage Weil <sage@newdream.net>
Looks like this was missed when flocklock was added. Did a quick grep and
it doesn't look like it is missing anywhere else.
Signed-off-by: Sage Weil <sage@newdream.net>
Otherwise we don't actually journal this value, and we get confused when
we replay a start_truncate and try to restart it.
Fixes: #1756
Signed-off-by: Sage Weil <sage.weil@dreamhost.com>
These flags are probably relics from when the function got split;
they belong in send_op now.
Signed-off-by: Greg Farnum <gregory.farnum@dreamhost.com>
If an address isn't explicitly specified during mkfs, look for an unnamed
monitor in the (generated) monmap and see if any of those addresses is
configured on the local machine. If so, assume it's us, and name ourselves
in the seed monmap.
Signed-off-by: Sage Weil <sage@newdream.net>
Previously, the overlaps for snapdirs would not be included in
cstat causing the computed total to be incorrect.
Signed-off-by: Samuel Just <samuel.just@dreamhost.com>
