summaryrefslogtreecommitdiff
Commit message (Collapse)AuthorAgeFilesLines
* rgw: don't iteraste through all objects when in namespacewip-4363Yehuda Sadeh2013-03-071-1/+21
| | | | | | | | | | Fixes: #4363 Backport: argnaut, bobtail When listing objects in namespace don't iterate through all the objects, only go though the ones that starts with the namespace prefix Signed-off-by: Yehuda Sadeh <yehuda@inktank.com>
* rgw: a tool to fix clobbered bucket info in user's bucket listYehuda Sadeh2013-02-071-1/+48
| | | | | | | | | | This fixes bad entries in user's bucket list that may have occured due to issue #4039. Syntax: $ radosgw-admin user check --uid=<uid> [--fix] Signed-off-by: Yehuda Sadeh <yehuda@inktank.com> Reviewed-by: Greg Farnum <greg@inktank.com>
* rgw: bucket recreation should not clobber bucket infoYehuda Sadeh2013-02-072-1/+24
| | | | | | | | | | Fixes: #4039 User's list of buckets is getting modified even if bucket already exists. This fix removes the newly created directory object, and makes sure that user info's data points at the correct bucket. Signed-off-by: Yehuda Sadeh <yehuda@inktank.com> Reviewed-by: Greg Farnum <greg@inktank.com>
* rgw: unlink multipart upload parts when completing uploadYehuda Sadeh2013-02-051-2/+13
| | | | | | | | | | | Fixes: #4011 When completing the multipart upload, we also need to unlink the parts from the bucket index. Originally we used to remove the parts however, nowadays the parts live on as we just point the object manifest at them. So we don't remove the objects, however, we need to remove them from the bucket index. Signed-off-by: Yehuda Sadeh <yehuda@inktank.com>
* rgw: a tool to fix buckets with leaked multipart referencesYehuda Sadeh2013-02-053-9/+109
| | | | | | | | | | Checks specified bucket for the #4011 symptoms, optionally fix the issue. sytax: radosgw-admin bucket check --bucket=<bucket> [--fix] Signed-off-by: Yehuda Sadeh <yehuda@inktank.com>
* rgw: radosgw-admin object unlinkYehuda Sadeh2013-02-054-0/+43
| | | | | | Add a radosgw-admin option to remove object from bucket index Signed-off-by: Yehuda Sadeh <yehuda@inktank.com>
* Merge remote-tracking branch 'gh/wip-scrub-argonaut' into argonautSage Weil2013-01-188-45/+227
|\
| * osd: send forced scrub/repair through scrub schedulingSage Weil2013-01-154-20/+34
| | | | | | | | | | | | | | | | | | | | | | | | | | | | This marks a PG for immediate scrub or repair. Adjust the sched_scrub() code so that we handle these PGs even when should_schedule_scrub is false (e.g., because the load is high). When we explicitly request a scrub or repair, we then go through the normal scrub reservation process to avoid unduly impacting cluster performance. This is particularly helpful on argonaut, where the final scrub finalization step blocks writes to the PG, and overlapping scrubs can exacerbate the problem. Signed-off-by: Sage Weil <sage@inktank.com>
| * osd: use helpers to queue a PG in the scrub LRUSage Weil2013-01-153-14/+25
| | | | | | | | | | | | | | Move the duplicated reach into info.history.last_scrub_stamp into a helper so we can control when we queue the PG for scrub. Signed-off-by: Sage Weil <sage@inktank.com>
| * osd/ReplicatedPG: validate ino when scrubbing snap collectionsSage Weil2013-01-154-4/+7
| | | | | | | | Signed-off-by: Sage Weil <sage@inktank.com>
| * ReplicatedPG: compare nlinks to snapcollsSamuel Just2013-01-151-0/+6
| | | | | | | | | | | | | | | | | | nlinks gives us the number of hardlinks to the object. nlinks should be 1 + snapcolls.size(). This will allow us to detect links which remain in an erroneous snap collection. Signed-off-by: Samuel Just <sam.just@inktank.com>
| * ReplicatedPG/PG: check snap collections during _scan_listSamuel Just2013-01-154-2/+119
| | | | | | | | | | | | | | | | | | During _scan_list check the snapcollections corresponding to the object_info attr on the object. Report inconsistencies during scrub_finalize. Signed-off-by: Samuel Just <sam.just@inktank.com> Reviewed-by: Sage Weil <sage@inktank.com>
| * byteorder: fix gcc 4.7 warningsSage Weil2013-01-151-3/+3
| | | | | | | | | | | | | | ./include/encoding.h: In function 'void encode(int64_t, ceph::bufferlist&, uint64_t)': ./include/encoding.h:101:1: warning: narrowing conversion of 'v' from 'int64_t {aka long int}' to '__le64 {aka long long unsigned int}' inside { } is ill-formed in C++11 [-Wnarrowing] Signed-off-by: Sage Weil <sage@inktank.com>
| * osd_types: add nlink and snapcolls fields to ScrubMap::objectSamuel Just2013-01-152-3/+17
| | | | | | | | Signed-off-by: Samuel Just <sam.just@inktank.com>
| * osd_types: bring ScrubMap::object up to the 0.56.1 encodingSamuel Just2013-01-152-3/+15
| | | | | | | | | | | | | | We need to introduce some new fields here, so to maintain compatibility we'll need to first bring the 48.* series up to the current encoding. Signed-off-by: Samuel Just <sam.just@inktank.com>
| * osd: make missing head non-fatal during scrubSage Weil2013-01-151-2/+7
| | | | | | | | | | | | | | | | | | If we encounter a scrub without a preceeding head, warn instead of crashing. Note that this is still something we can't repair. See #3705. Signed-off-by: Sage Weil <sage@inktank.com>
* | os/: Add CollectionIndex::prep_deleteSamuel Just2013-01-184-1/+48
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | If an unlink is interupted between removing the file and updating the subdir attribute, the attribute will overestimate the number of files in the directory. This is by design, at worst we will merge the collection later than intended, but closing the gap would require a second subdir xattr update. However, this can in extreme cases result in a collection with subdirectories but no objects. FileStore::_destry_collection would therefore see an erroneous -ENOTEMPTY. prep_delete allows the CollectionIndex implementation to clean up state prior to removal. Signed-off-by: Samuel Just <sam.just@inktank.com> (cherry picked from commit fdc5e5d1877d7d7ed3851b9ec01f884559748249) Conflicts: src/os/HashIndex.cc src/os/HashIndex.h
* | rgw: copy object should not copy source aclsYehuda Sadeh2013-01-161-0/+3
|/ | | | | | | | | | | | | | | Fixes: #3802 Backport: argonaut, bobtail When using the S3 api and x-amz-metadata-directive is set to COPY we used to copy complete metadata of source object. However, this shouldn't include the source ACLs. Conflicts: src/rgw/rgw_rados.cc Signed-off-by: Yehuda Sadeh <yehuda@inktank.com> (cherry picked from commit ccfefe3097a51b49885f2ed5d9334e85b497d963)
* ReplicatedPG: fix snapdir trimmingSamuel Just2013-01-101-9/+28
| | | | | | | | | | | | The previous logic was both complicated and not correct. Consequently, we have been tending to drop snapcollection links in some cases. This has resulted in clones incorrectly not being trimmed. This patch replaces the logic with something less efficient but hopefully a bit clearer. Signed-off-by: Samuel Just <sam.just@inktank.com> Reviewed-by: Sage Weil <sage@inktank.com> (cherry picked from commit 0f42c37359d976d1fe90f2d3b877b9b0268adc0b)
* v0.48.3argonautv0.48.3argonautGary Lowell2013-01-072-2/+8
|
* osd: fix race in do_recovery()Sage Weil2013-01-072-0/+9
| | | | | | | | | | Verify that the PG is still RECOVERING or BACKFILL when we take the pg lock in the recovery thread. This prevents a crash from an invalid state machine event when the recovery queue races with a PG state change (e.g., due to peering). Signed-off-by: Sage Weil <sage@inktank.com> Reviewed-by: Samuel Just <sam.just@inktank.com>
* ReplicatedPG: requeue waiting_for_ondisk in apply_and_flush_repopsSamuel Just2013-01-071-12/+19
| | | | | | Fixes: #3722 Signed-off-by: Samuel Just <sam.just@inktank.com> Reviewed-by: Sage Weil <sage@inktank.com>
* ObjectCacher: fix off-by-one error in splitJosh Durgin2013-01-041-9/+10
| | | | | | | | | | | | | | | | | | | | | | | | This error left a completion that should have been attached to the right BufferHead on the left BufferHead, which would result in the completion never being called unless the buffers were merged before it's original read completed. This would cause a hang in any higher level waiting for a read to complete. The existing loop went backwards (using a forward iterator), but stopped when the iterator reached the beginning of the map, or when a waiter belonged to the left BufferHead. If the first list of waiters should have been moved to the right BufferHead, it was skipped because at that point the iterator was at the beginning of the map, which was the main condition of the loop. Restructure the waiters-moving loop to go forward in the map instead, so it's harder to make an off-by-one error. Possibly-fixes: #3286 Signed-off-by: Josh Durgin <josh.durgin@inktank.com> (cherry picked from commit 2e862f4d183d8b57b43b0777737886f18f68bf00)
* config: change default log_max_recent to 10,000Sage Weil2013-01-041-1/+1
| | | | | | | Commit c34e38bcdc0460219d19b21ca7a0554adf7f7f84 meant to do this but got the wrong number of zeros. Signed-off-by: Sage Weil <sage@inktank.com>
* os/FileStore: fix non-btrfs op_seq commit orderSage Weil2013-01-031-5/+16
| | | | | | | | | | | | | | | | | | | The op_seq file is the starting point for journal replay. For stable btrfs commit mode, which is using a snapshot as a reference, we should write this file before we take the snap. We normally ignore current/ contents anyway. On non-btrfs file systems, however, we should only write this file *after* we do a full sync, and we should then fsync(2) it before we continue (and potentially trim anything from the journal). This fixes a serious bug that could cause data loss and corruption after a power loss event. For a 'kill -9' or crash, however, there was little risk, since the writes were still captured by the host's cache. Fixes: #3721 Signed-off-by: Sage Weil <sage@inktank.com> Reviewed-by: Samuel Just <sam.just@inktank.com> (cherry picked from commit 28d59d374b28629a230d36b93e60a8474c902aa5)
* log: broadcast cond signalsSage Weil2013-01-032-8/+14
| | | | | | | | | | | | | | | | | | | | We were using a single cond, and only signalling one waiter. That means that if the flusher and several logging threads are waiting, and we hit a limit, we the logger could signal another logger instead of the flusher, and we could deadlock. Similarly, if the flusher empties the queue, it might signal only a single logger, and that logger could re-signal the flusher, and the other logger could wait forever. Intead, break the single cond into two: one for loggers, and one for the flusher. Always signal the (one) flusher, and always broadcast to all loggers. Backport: bobtail, argonaut Signed-off-by: Sage Weil <sage@inktank.com> Reviewed-by: Dan Mick <dan.mick@inktank.com> (cherry picked from commit 813787af3dbb99e42f481af670c4bb0e254e4432)
* log: fix locking typo/stupid for dump_recent()Sage Weil2013-01-031-1/+1
| | | | | | | | | | We weren't locking m_flush_mutex properly, which in turn was leading to racing threads calling dump_recent() and garbling the crash dump output. Backport: bobtail, argonaut Signed-off-by: Sage Weil <sage@inktank.com> Reviewed-by: Dan Mick <dan.mick@inktank.com> (cherry picked from commit 43cba617aa0247d714632bddf31b9271ef3a1b50)
* init-ceph: fix status version check across machinesSage Weil2012-12-281-2/+2
| | | | | | | | The local state isn't propagated into the backtick shell, resulting in 'unknown' for all remote daemons. Avoid backticks altogether. Signed-off-by: Sage Weil <sage@inktank.com> (cherry picked from commit 6c7b667badc5e7608b69c533a119a2afc062e257)
* init-ceph: use SSH in "service ceph status -a" to get versionTravis Rhoden2012-12-281-1/+1
| | | | | | | | | | | | | | When running "service ceph status -a", a version number was never returned for remote hosts, only for the local. This was because the command to query the version number didn't use the do_cmd function, which is responsible for running the command over SSH when needed. Modify the ceph init.d script to use do_cmd for querying the Ceph version. Signed-off-by: Travis Rhoden <trhoden@gmail.com> (cherry picked from commit 60fdb6fda6233b01dae4ed8a34427d5960840b84)
* log: 10,000 recent log entriesSage Weil2012-12-261-1/+1
| | | | | | | | This is what we were (wrongly) doing before, so there are no memory utilization surprises. Signed-off-by: Sage Weil <sage@inktank.com> (cherry picked from commit 78286b1403a5e0f14f95fe6b92f2fdb163e909f1)
* log: fix log_max_recent configSage Weil2012-12-261-1/+1
| | | | | | | <facepalm> Signed-off-by: Sage Weil <sage@inktank.com> (cherry picked from commit 4de7748b72d4f90eb1197a70015c199c15203354)
* log: fix flush/signal raceSage Weil2012-12-261-1/+1
| | | | | | | | | | | | | | | | | We need to signal the cond in the same interval where we hold the lock *and* modify the queue. Otherwise, we can have a race like: queue has 1 item, max is 1. A: enter submit_entry, signal cond, wait on condition B: enter submit_entry, signal cond, wait on condition C: flush wakes up, flushes 1 previous item A: retakes lock, enqueues something, exits B: retakes lock, condition fails, waits -> C is never woken up as there are 2 items waiting Signed-off-by: Sage Weil <sage@inktank.com> Reviewed-by: Dan Mick <dan.mick@inktank.com> (cherry picked from commit 50914e7a429acddb981bc3344f51a793280704e6)
* .gitignore: Add ar-lib to ignore listGary Lowell2012-12-211-0/+1
|
* autogen.sh: Create m4 directory for leveldbGary Lowell2012-12-211-1/+1
| | | | Signed-off-by: Gary Lowell <gary.lowell@inktank.com>
* leveldb: Update submoduleGary Lowell2012-12-211-0/+0
| | | | Signed-off-by: Gary Lowell <gary.lowell@inktank.com>
* ceph.spec.in: Fedora builds debuginfo by default.Gary Lowell2012-12-201-1/+1
| | | | Signed-off-by: Gary Lowell <gary.lowell@inktank.com>
* rgw: fix error handling with swiftYehuda Sadeh2012-12-191-1/+1
| | | | | | | Fixes: #3649 verify_swift_token returns a bool and not an int. Signed-off-by: Yehuda Sadeh <yehuda@inktank.com>
* client: Fix for #3184 cfuse segv with no keyringSam Lang2012-12-112-8/+25
| | | | | | | | | Fixes bug #3184 where the ceph-fuse client segfaults if authx is enabled but no keyring file is present. This was due to the client->init() return value not getting checked. Signed-off-by: Sam Lang <sam.lang@inktank.com> (cherry picked from commit 47983df4cbd31f299eef896b4612d3837bd7c7bd)
* mon: Monitor: resolve keyring option to a file before loading keyringJoao Eduardo Luis2012-12-111-1/+7
| | | | | | | | | | | | Otherwise our keyring default location, or any other similarly formatted location, will be taken as the actual location for the keyring and fail. Reported-by: tziOm (at) #ceph Fixes: 3276 Backport: argonaut Signed-off-by: Joao Eduardo Luis <joao.luis@inktank.com> (cherry picked from commit 7ef0df25e001bfae303feb3ae36514608767b1f2)
* .gitignore: Add m4 macro directories to ignore listGary Lowell2012-12-051-1/+5
|
* build: Add RPM release string generated from git describe.Gary Lowell2012-12-032-1/+7
| | | | | | | | | Fix for bug 3451. Use the commit count and sha1 from git describe to construct a release string for rpm packages. Conflicts: configure.ac
* ceph.spec.in: Build debuginfo subpackage.Gary Lowell2012-12-031-0/+7
| | | | | | This is a partial fix for bug 3471. Enable building of debuginfo package. Some distributions enable this automatically by installing additional rpm macros, on others it needs to be explicity added to the spec file.
* rgw: fix swift auth concurrency issueYehuda Sadeh2012-12-031-1/+1
| | | | | | | | | Fixes: #3565 Originally ops were using static structures, but that has since changed. Switching swift auth handler to do the same. Signed-off-by: Yehuda Sadeh <yehuda@inktank.com>
* rgw: fix rgw_tools get_obj()Yehuda Sadeh2012-12-032-10/+14
| | | | | | | | | The original implementation broke whenever data exceeded the chunk size. Also don't keep cache for objects that exceed the chunk size as cache is not designed for it. Increased chunk size to 512k. Signed-off-by: Yehuda Sadeh <yehuda@inktank.com>
* rgw: fix PUT aclsYehuda Sadeh2012-11-291-7/+10
| | | | | | | | | This fixes a regression introduced at 17e4c0df44781f5ff1d74f3800722452b6a0fc58. The original patch fixed error leak, however it also removed the operation's send_response() call. Signed-off-by: Yehuda Sadeh <yehuda@inktank.com>
* rgw: fix xml parser leakYehuda Sadeh2012-11-211-4/+3
| | | | | Signed-off-by: Yehuda Sadeh <yehuda@inktank.com> (cherry picked from commit f86522cdfcd81b2d28c581ac8b8de6226bc8d1a4)
* rgw: fix memory leaksYehuda Sadeh2012-11-212-15/+18
| | | | | | | | | Signed-off-by: Yehuda Sadeh <yehuda@inktank.com> (cherry picked from commit 98a04d76ebffa61c3ba4b033cdd57ac57b2f29f3) Conflicts: src/rgw/rgw_op.cc src/rgw/rgw_op.h
* rgw: don't convert object mtime to UTCYehuda Sadeh2012-11-151-8/+1
| | | | | | | | Fixes: #3452 When we read object info, don't try to convert mtime to UTC, it's already in UTC. Signed-off-by: Yehuda Sadeh <yehuda@inktank.com>
* rgw: relax date format checkYehuda Sadeh2012-11-141-8/+0
| | | | | | | | | Don't try to parse beyond the GMT or UTC. Some clients use special date formatting. If we end up misparsing the date it'll fail in the authorization, so don't need to be too restrictive. Signed-off-by: Yehuda Sadeh <yehuda@inktank.com>
* ceph-disk-activate: avoid duplicating mounts if already activatedSage Weil2012-11-121-5/+20
| | | | | | | | | If the given device is already mounted at the target location, do not mount --move it again and create a bunch of dup entries in the /etc/mtab and kernel mount table. Signed-off-by: Sage Weil <sage@inktank.com> (cherry picked from commit c435d314caeb5424c1f4482ad02f8a085317ad5b)
View Lorry Controller Status