ceph-disk-activate: identify cluster .conf by fsid

Determine what cluster the disk belongs to by checking the fsid defined
in /etc/ceph/*.conf.  Previously we hard-coded 'ceph'.

Note that this has the nice side-effect that if we have a disk with a
bad/different fsid, we now fail to activate it.  Previously, we would
mount and start ceph-osd, but the daemon would fail to authenticate
because it was part of the wrong cluster.

Fixes: #3253
Signed-off-by: Sage Weil <sage@inktank.com>

ceph-disk-activate: abort if target position is already mounted

If the target position is already a mount point, fail to move our mount
over to it.  This usually indicates that a different osd.N from a
different cluster instances is in that position.

Signed-off-by: Sage Weil <sage@inktank.com>

rados/test.sh fails in the nightly run

Make test more robust by using my_snaps vector for snap IDs

Signed-off-by: David Zafman <david.zafman@inktank.com>

Merge remote-tracking branch 'origin/wip-osd-addr-api'

Reviewed-by: Sage Weil <sage@inktank.com>

debian: add start ceph-mds-all on ceph-mds install

This ensures that when we then start individual mds instances, we can
stop ceph-mds-all and they will get stopped.  We do the same already for
ceph-all.

Signed-off-by: Sage Weil <sage@inktank.com>

libcephfs: add ceph_get_osd_addr interface

Return the network address for an OSD by ID.

Signed-off-by: Noah Watkins <noahwatkins@gmail.com>

Revert "ceph-disk-activate: rely on default/configured keyring path"

This reverts commit 936b8f20af1d390976097c427b6e92da4b39b218.

This is necessary because we mount the osd in a temporary location.

Signed-off-by: Sage Weil <sage@inktank.com>

Revert "ceph-disk-activate: don't override default or configured osd journal path"

This reverts commit 813e9fe2b4291a1c1922ef78f031daa9b78fe53b.

We run --mkfs with the osd disk mounted in a temporary location, so it is
necessary to explicitly pass in these paths.

If we want to support journals in a different location, we need to make
ceph-disk-prepare update the journal symlink accordingly.. not control it via
the config option.

Signed-off-by: Sage Weil <sage@inktank.com>

Merge pull request #109 from dalgaaf/wip-da-performance-1-v2

prefer prefix ++/--operator for e.g. iterators for performance reasons

Reviewed-by: Sage Weil <sage@inktank.com>

Merge pull request #108 from ceph/wip-refuse-last-mon-remove

mon: refuse "mon remove" if only one mon left

Reviewed-by: Sage Weil <sage@inktank.com>

monmaptool.cc: prefer prefix ++operator for iterators

Signed-off-by: Danny Al-Gaaf <danny.al-gaaf@bisect.de>

mon/PGMonitor.cc: prefer prefix ++operator for iterators

Signed-off-by: Danny Al-Gaaf <danny.al-gaaf@bisect.de>

mon/MonmapMonitor.cc: prefer prefix ++operator for iterators

Signed-off-by: Danny Al-Gaaf <danny.al-gaaf@bisect.de>

mon/Monitor.cc: prefer prefix ++operator for iterators

Signed-off-by: Danny Al-Gaaf <danny.al-gaaf@bisect.de>

mon/MonMap.cc: prefer prefix ++operator for iterators

Signed-off-by: Danny Al-Gaaf <danny.al-gaaf@bisect.de>

mon/MDSMonitor.cc: prefer prefix ++operator for iterators

Signed-off-by: Danny Al-Gaaf <danny.al-gaaf@bisect.de>

mon/LogMonitor.cc: prefer prefix ++operator for iterators

Signed-off-by: Danny Al-Gaaf <danny.al-gaaf@bisect.de>

mon/AuthMonitor.cc: prefer prefix ++operator for iterators

Signed-off-by: Danny Al-Gaaf <danny.al-gaaf@bisect.de>

mds/mdstypes.cc: prefer prefix ++operator for iterators

Signed-off-by: Danny Al-Gaaf <danny.al-gaaf@bisect.de>

mds/SnapServer.cc: prefer prefix ++operator for iterators

Signed-off-by: Danny Al-Gaaf <danny.al-gaaf@bisect.de>

mds/SnapServer.h: prefer prefix ++operator for iterators

Signed-off-by: Danny Al-Gaaf <danny.al-gaaf@bisect.de>

mds/SnapServer.h: prefer prefix ++operator for iterators

Signed-off-by: Danny Al-Gaaf <danny.al-gaaf@bisect.de>

mds/Mutation.cc: prefer prefix ++operator for iterators

Signed-off-by: Danny Al-Gaaf <danny.al-gaaf@bisect.de>

mds/MDSTableServer.cc: prefer prefix ++operator for iterators

Signed-off-by: Danny Al-Gaaf <danny.al-gaaf@bisect.de>

mds/MDSTableClient.cc: prefer prefix ++operator for iterators

Signed-off-by: Danny Al-Gaaf <danny.al-gaaf@bisect.de>

mds/MDSMap.cc: prefer prefix ++operator for iterators

Signed-off-by: Danny Al-Gaaf <danny.al-gaaf@bisect.de>

mds/MDS.cc: prefer prefix ++operator for iterators

Signed-off-by: Danny Al-Gaaf <danny.al-gaaf@bisect.de>

mds/MDLog.cc: prefer prefix ++operator for iterators

Signed-off-by: Danny Al-Gaaf <danny.al-gaaf@bisect.de>

mds/MDCache.cc: prefer prefix ++operator for iterators

Signed-off-by: Danny Al-Gaaf <danny.al-gaaf@bisect.de>

mds/MDBalancer.cc: prefer prefix ++operator for iterators

Signed-off-by: Danny Al-Gaaf <danny.al-gaaf@bisect.de>

mds/Locker.cc: prefer prefix ++operator for iterators

Signed-off-by: Danny Al-Gaaf <danny.al-gaaf@bisect.de>

mds/MDCache.h: prefer prefix ++operator for iterators

Signed-off-by: Danny Al-Gaaf <danny.al-gaaf@bisect.de>

mds/CInode.cc: prefer prefix ++operator for iterators

Signed-off-by: Danny Al-Gaaf <danny.al-gaaf@bisect.de>

mds/CDir.cc: prefer prefix ++operator for iterators

Signed-off-by: Danny Al-Gaaf <danny.al-gaaf@bisect.de>

mds/AnchorServer.h: prefer prefix ++operator for iterators

Signed-off-by: Danny Al-Gaaf <danny.al-gaaf@bisect.de>

mds/AnchorServer.cc: prefer prefix ++operator for iterators

Signed-off-by: Danny Al-Gaaf <danny.al-gaaf@bisect.de>

mds/AnchorClient.cc: prefer prefix ++operator for iterators

Signed-off-by: Danny Al-Gaaf <danny.al-gaaf@bisect.de>

librbd/internal.cc: prefer prefix ++operator for iterators

Prefer prefix ++operator for iterators, move iterator declaration into
the related for loop.

Signed-off-by: Danny Al-Gaaf <danny.al-gaaf@bisect.de>

librbd/ImageCtx.cc: prefer prefix ++operator for iterators

Signed-off-by: Danny Al-Gaaf <danny.al-gaaf@bisect.de>

librados/RadosClient.cc: prefer prefix ++operator for iterators

Signed-off-by: Danny Al-Gaaf <danny.al-gaaf@bisect.de>

librados/IoCtxImpl.cc: prefer prefix ++operator for iterators

Signed-off-by: Danny Al-Gaaf <danny.al-gaaf@bisect.de>

libcephfs.cc: prefer prefix ++operator for iterators

Signed-off-by: Danny Al-Gaaf <danny.al-gaaf@bisect.de>

kv_flat_btree_async.cc: prefer prefix ++operator for iterators

Signed-off-by: Danny Al-Gaaf <danny.al-gaaf@bisect.de>

key_value_store/cls_kvs.cc: prefer prefix --operator for iterators

Signed-off-by: Danny Al-Gaaf <danny.al-gaaf@bisect.de>

crush/CrushWrapper.cc: prefer prefix ++operator for iterators

Signed-off-by: Danny Al-Gaaf <danny.al-gaaf@bisect.de>

crush/CrushCompiler.cc: prefer prefix --operator for iterators

Signed-off-by: Danny Al-Gaaf <danny.al-gaaf@bisect.de>

common/lockdep.cc: prefer prefix ++operator for iterators

Signed-off-by: Danny Al-Gaaf <danny.al-gaaf@bisect.de>

client/SyntheticClient.cc: prefer prefix ++operator for iterators

Signed-off-by: Danny Al-Gaaf <danny.al-gaaf@bisect.de>

mon: refuse "mon remove" if only one mon left

Fixes: #4439
Signed-off-by: Dan Mick <dan.mick@inktank.com>

osd: data loss: low space handling

Add check whether to allow writing ops based on failsafe full percentage
Check for failsafe nearfull warning or full error message every heartbeat
Use clock to limit messages to every 30 secs (osd_op_complaint_time)

Feature: #4197

Signed-off-by: David Zafman <david.zafman@inktank.com>
Reviewed-by: Samuel Just <sam.just@inktank.com>

Simplify may_write()

CEPH_MDS_OP_CREATE has CEPH_MDS_OP_WRITE bit set so already checked

Signed-off-by: David Zafman <david.zafman@inktank.com>
Reviewed-by: Samuel Just <sam.just@inktank.com>

Merge branch 'wip_omap_snaps'

Reviewed-by: Greg Farnum <greg@inktank.com>

OSD: remove snap_mapper entries in RemoveWQ

This is a somewhat ugly hack for now until we implement
pg resurrection.

Signed-off-by: Samuel Just <sam.just@inktank.com>

PG: check and repair snap_mapper entries during scrub

Signed-off-by: Samuel Just <sam.just@inktank.com>

OSD,PG: add upgrade procedure for snap_mapper

Also, sub_op_modify transactions currently carry the operations
for creating snap links in the shipped transaction.  To handle
ops shipped by unenlightened osds, transactions can now be
tagged with a tolerate_collection_add_enoent flag.

Signed-off-by: Samuel Just <sam.just@inktank.com>

osd/: Integrate SnapMapper with OSD

- SnapTrimmer now uses SnapMapper to get the next object to trim
- Entries for a snap are implicitely removed from SnapMapper when
the last object is trimmed, so no need for the adjust_local_snaps
logic.
- Scrub now compares the object_info snaps set on the object attr
with the version stored in the SnapMapper.

Signed-off-by: Samuel Just <sam.just@inktank.com>

Makefile: add SnapMapper to osd

Signed-off-by: Samuel Just <sam.just@inktank.com>

test/: add tests for map_cacher

Signed-off-by: Samuel Just <sam.just@inktank.com>

osd/: add SnapMapper

Signed-off-by: Samuel Just <sam.just@inktank.com>

common/: add map_cacher

Signed-off-by: Samuel Just <sam.just@inktank.com>

common/shared_ptr.hpp: add get_next and expose VPtr and WeakVPtr types

Signed-off-by: Samuel Just <sam.just@inktank.com>

filestore: add debug option to verify split results

Signed-off-by: Samuel Just <sam.just@inktank.com>

DBObjectMap, hobject: add helpers for pgid bit matching

Create helpers in hobject for generating prefixes for a
pg as well as matching hobjects against a pgid/numpgs
combo.  Use these in HashIndex.cc.

Signed-off-by: Samuel Just <sam.just@inktank.com>

ObjectStore: add Context management to Transaction

ObjectStore now appends passed contexts in queue_transaction
to the Transaction contexts and uses that to pass into
the virtual queue_transactions.

Signed-off-by: Samuel Just <sam.just@inktank.com>

Context: allow C_Contexts to not have a cct, add list_to_context

This will simplify the SnapMapper implementation.

Signed-off-by: Samuel Just <sam.just@inktank.com>

ObjectStore: move queue_transaction into ObjectStore

This will simplify the SnapMapper implementation.

Signed-off-by: Samuel Just <sam.just@inktank.com>

osd_internals/snaps.rst: update description to reflect SnapMapper

Signed-off-by: Samuel Just <sam.just@inktank.com>

osd_internals/snaps.rst: add a description of snaps and trimming

Signed-off-by: Samuel Just <sam.just@inktank.com>

PG::read_log: fix assert, split may introduce holes in the log

Signed-off-by: Samuel Just <sam.just@inktank.com>

OSD: add debugging to start_split and complete_split

Signed-off-by: Samuel Just <sam.just@inktank.com>

PG: check_recovery_sources must happen even if not active

missing_loc/missing_loc_sources also must be cleaned up
if a peer goes down during peering:

1) pg is in GetInfo, acting is [3,1]
2) we find object A on osd [0] in GetInfo
3) 0 goes down, no new peering interval since it is neither up nor
acting, but peer_missing[0] is removed.
4) pg goes active and try to pull A from 0 since missing_loc did not get
cleaned up.

Backport: bobtail
Fixes: #4371
Signed-off-by: Samuel Just <sam.just@inktank.com>
Reviewed-by: Greg Farnum <greg@inktank.com>

FileStore: fix reversed collection_empty return value

Backport: bobtail
Fixes: #4380
Signed-off-by: Samuel Just <sam.just@inktank.com>
Reviewed-by: Greg Farnum <greg@inktank.com>

HashIndex: _collection_list_partial must tolerate NULL next

Backport: bobtail
Fixes: #4379
Signed-off-by: Samuel Just <sam.just@inktank.com>
Reviewed-by: Greg Farnum <greg@inktank.com>

OSD: lock not needed in ~DeletingState()

No further refs to the object can remain at this point.
Furthermore, the callbacks might lock mutexes of their
own.

Backport: bobtail
Fixes: #4378
Signed-off-by: Samuel Just <sam.just@inktank.com>
Reviewed-by: Greg Farnum <greg@inktank.com>

ReplicatedPG: don't leak reservation on removal

Fixes: 4431
Signed-off-by: Samuel Just <sam.just@inktank.com>
Reviewed-by: Greg Farnum <greg@inktank.com>

Merge remote-tracking branch 'dalgaaf/wip-da-performance'

Conflicts:
src/client/Client.cc

Reviewed-by: Sage Weil <sage@inktank.com>

Merge branch 'wip-mds-con'

Reviewed-by: Greg Farnum <greg@inktank.com>
Tested-by: Noah Watkins <jayhawk@cs.ucsc.edu>

mds: log session resets more cleanly

Signed-off-by: Sage Weil <sage@inktank.com>

mds: use session con ref to mark down killing session

Signed-off-by: Sage Weil <sage@inktank.com>

mds: mark con for closed session disposable

If there is a fault while delivering the message, close the con.  This will
clean up the Session state from memory.  If the client doesn't get the
CLOSED message, they will reconnect (from their perspective, it is still
a lossless connection) and get a remote_reset event telling them that the
session is gone.  The client code already handles this case properly.

Note that way back in 4ac45200f10e0409121948cea5226ca9e23bb5fb we removed
this because the client would reuse the same connection when it reopened
the session.  Now the client never does that; it will mark_down the con
as soon as it is closed and open a new one for a new session... which means
the MDS will get a remote_reset and close out the old session.

Signed-off-by: Sage Weil <sage@inktank.com>

mds: useless cleanup

Signed-off-by: Sage Weil <sage@inktank.com>

mds: mark_down old sessions using con

Use the Connection ref to mark old client sessions down.  This is cleaner.

Signed-off-by: Sage Weil <sage@inktank.com>

Merge branch 'wip-4390-b'

Reviewed-by: Greg Farnum <greg@inktank.com>

client: fix kick_flushing_caps() args

Signed-off-by: Sage Weil <sage@inktank.com>

client: clean up useless MetaSession lookups

Signed-off-by: Sage Weil <sage@inktank.com>

client: fix trim_caps() args

Signed-off-by: Sage Weil <sage@inktank.com>

client: fix wake_inode_waiters() arg

Signed-off-by: Sage Weil <sage@inktank.com>

client: validate/lookup mds session in each message handler

For every message handler, look up the MetaSession by int mds and verify
that the Connection* matches properly.  If so, proceed; otherwise, discard
the message.

In the future, we probably want to link the MetaSession to the Connection's
priv field, but that can come later.

Clean up a bunch of submethods that take int mds while we're here.

Signed-off-by: Sage Weil <sage@inktank.com>

client: fix remaining s->inst users to use s->con

Signed-off-by: Sage Weil <sage@inktank.com>

client: remove unused MetaSession::was_stale

Signed-off-by: Sage Weil <sage@inktank.com>

client: only signal session waiters on open and close

Signed-off-by: Sage Weil <sage@inktank.com>

client: use Connection* in MetaSession instead of the inst/addr

This moves us to a cleaner API, and makes the session lifetime explicit
and clean.  Yay!

Signed-off-by: Sage Weil <sage@inktank.com>

client: use session->inst where possible

Removing remaining lookups up the mds inst in the MDSMap.  Instead, use
the MetaSession inst field.

Signed-off-by: Sage Weil <sage@inktank.com>

client: drop unused release_lease()

Signed-off-by: Sage Weil <sage@inktank.com>

client: pass session to send_reconnect and resend_unsafe_requests

Signed-off-by: Sage Weil <sage@inktank.com>

client: switch kick_requests and send_request to take a session *

More cleanup!

Signed-off-by: Sage Weil <sage@inktank.com>

client: handle ESTALE redirection in make_request(), not reply handler

Resending the request in the reply handler is a bit fugly and throws a
small wrench into moving to a MetaSession*-based approach.  Check for
the case(s) where we *do* return ESTALE explicitly and fall through.
Otherwise, kick the caller and let them retry.

Signed-off-by: Sage Weil <sage@inktank.com>

client: fix up _closed_mds_session

Drop useless arg.  Simplify callers.

Signed-off-by: Sage Weil <sage@inktank.com>

client: use helper for _close_mds_session()

Signed-off-by: Sage Weil <sage@inktank.com>

mds: flush log when we hit xlock during path traversal

This ensures we don't stall out waiting for a lock state to change.

This fixes ~4-5 second stalls easily reproducible and visible with
ceph-fuse and 'dbench 1'.

Signed-off-by: Sage Weil <sage@inktank.com>