| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
Incorporate the new API 'gf_changelog_init'
introduced in libgfchangelog which should
be called before 'gf_changelog_register'.
Change-Id: I10a9dd5ff3ecc64fd61dcb6b1aa53739e0bdeedd
BUG: 1238063
Signed-off-by: Kotresh HR <khiremat@redhat.com>
Reviewed-on: http://review.gluster.org/11479
Tested-by: NetBSD Build System <jenkins@build.gluster.org>
Reviewed-by: Venky Shankar <vshankar@redhat.com>
Tested-by: Gluster Build System <jenkins@build.gluster.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
[WIP patch as of now, just needs a little tweak]
A pending TODO in the code caused regressions to fail as
bitrot daemons are spawned during volume start (equivalent
to enabling bitrot by default). The problematic part that
casued such failures is during brick disconnections with
unsafe handling of event data structured in the code.
With this patch, data structures are properly cleaned up
with care taken to cleanup all accessors first. This also
fixes potential memory leaks which was bluntly ignored
before.
Change-Id: I70ed82cb1a0fb56c85ef390007e321a97a35c5ce
BUG: 1170075
Signed-off-by: Venky Shankar <vshankar@redhat.com>
original-author: Venky Shankar <vshankar@redhat.com>
Reviewed-on: http://review.gluster.org/9959
Reviewed-by: Vijay Bellur <vbellur@redhat.com>
Tested-by: Vijay Bellur <vbellur@redhat.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Problem:
libgfchangelog initializes global xlator on library load (via
constructor: _ctor) and mangles it's xlator context thereby
messing with certain important members of the command structure.
On receiving an RPC disconnection event, if the point-of-execution
was in libgfchangelogs context, accessing ->cmd_args during RPC
notify resulted in a segfault.
Fix:
Since the libarary needs to be able to work with processes that
have a notion of an xlator (THIS in particular) and without it,
care needs to be taken to allocate the global xlator when needed.
Moreover, the actual fix is to use the correct xlator context
in both cases. A new API is introduces when needs to be invoked
by the conusmer (although this could have been done during
register() call, keeping it a separate API makes thing flexible
and easy).
Test:
The issue is observed when a brick process goes offline. This is
triggered when test cases (.t's) are run in bulk, since each
test essestially spawns bricks processes (on volume start) and
terminates them (volume stop). Since bitrot daemon, as of now,
spawns upon volume start, the issue is much observed when the
volume is taken offline at the end of each test case. With this
fix, running the basic and core test cases along with building
the linux kernel has passed without daemon segfaults.
Thanks to Johnny (rabhat@) for helping in debugging the issue
(and with the fix :)).
Change-Id: I8d3022bf749590b2ee816504ed9b1dfccc65559a
BUG: 1170075
Signed-off-by: Venky Shankar <vshankar@redhat.com>
Reviewed-on: http://review.gluster.org/9953
Reviewed-by: Vijay Bellur <vbellur@redhat.com>
Tested-by: Vijay Bellur <vbellur@redhat.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This patch introduces RPC based communication between the changelog
translator and libgfchangelog. It replaces the old pathetic stream
based interaction that existed earlier (due to time constraints :-/).
Changelog, upon initialization starts a RPC server (rpcsvc) allowing
clients to invoke a probe API as a bootup mechanism to request for
event notifications. During probe, clients can choose an event
filter specifying the type(s) of events they are interested in. As
of now there is no way to change the event notification set once
the probe RPC call is made, but that is easier to implement.
The actual event notifications is done on a separate RPC session.
The client (libgfchangelog) itself starts and RPC server which the
changelog translator "connects back" during probe. Notifications
are dispatched by a bunch of threads from the server (translator)
and the client optionally orders them if ordered notifications
are requried. FOPs fill in their respective event details in a
buffer (rot-buffs to be particular) and a bunch of threads
(consumers) swap the buffers out of roatation and dispatch them
via RPC. To avoid writer starvation, then number of dispatcher
threads is one less than the number of buffer list in rot-buffs.x
libgfchangelog becomes purely callback based -- upon event
notification from the server (and re-ordering them if required)
invoke a callback routine specified by consumer(s).
A major part of the patch is also aimed at providing backward
compatibility for geo-replication, which was one of the main
consumer of the stream based API. Also, this patch does not\
"turn on" event notifications for all fops, just a bunch which
is currently in requirement. Another pain point is that the
server does not filter events before dispatching it to the
clients. That load is taken up by the client itself (although
it's done at the library layer rather than making it hard on
the callback implementor). This needs improvement and care
needs to be taken to not load the server up with expensive
filtering mechanisms.
Change-Id: Ibf60a432b68f2dfa60c6f9add2bcfd37a9c41395
BUG: 1170075
Signed-off-by: Venky Shankar <vshankar@redhat.com>
Reviewed-on: http://review.gluster.org/9708
Reviewed-by: Jeff Darcy <jdarcy@redhat.com>
Tested-by: Gluster Build System <jenkins@build.gluster.com>
Reviewed-by: Vijay Bellur <vbellur@redhat.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Facilitates Glusterfs with the ability to detect file-operations
happened in past by scanning the back-end(brick-level) glusterfs
journal (changelog).
Design:
* List of changelogs produces in one perfectly running session are
stored in htime file which also holds necessary information about
the session start and end time.
* Involves fixed sized seeks to identify N'th changelog in the list.
* Requires O(log n), (where n is number of changelogs in the list),
time to identify the end changelog for the given start-end time
interval.
Currently the background processing of changelogs is sub optimal. BZ
1097041 tracks the development effort.
For complete design, refer the below link:
http://lists.nongnu.org/archive/html/gluster-devel/2014-02/msg00206.html
Change-Id: I27e49f75e492e843084d0ecaf9130224d08462a0
BUG: 1091961
Signed-off-by: Ajeet Jha <ajha@redhat.com>
Signed-off-by: Venky Shankar <vshankar@redhat.com>
Signed-off-by: Ajeet Jha <ajha@redhat.com>
Reviewed-on: http://review.gluster.org/6930
Reviewed-by: Kotresh HR <khiremat@redhat.com>
Tested-by: Gluster Build System <jenkins@build.gluster.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
-> log additional records.
-> include FOP number for metadata.
-> prevent crash if inode is not found in a fop.
Change-Id: I9edd4b71819ebd68c6a2b4150ae279c471d129da
BUG: 1036536
Signed-off-by: Ajeet Jha <ajha@redhat.com>
Reviewed-on: http://review.gluster.org/6403
Tested-by: Gluster Build System <jenkins@build.gluster.com>
Reviewed-by: Amar Tumballi <amarts@gmail.com>
Reviewed-by: Venky Shankar <vshankar@redhat.com>
|
|
|
This is the initial version of the Changelog Translator.
What is it
-----------
Goal is to capture changes performed on a GlusterFS volume.
The translator needs to be loaded on the server (bricks) and
captures changes in a plain text file inside a configured
directory path (controlled by "changelog-dir", should be
somewhere in <export>/.glusterfs/changelog by default).
Changes are classified into 3 types:
- Data: : TYPE-I
- Metadata : TYPE-II
- Entry : TYPE-III
Changelog file is rolled over after a certain time interval
(defauls to 60 seconds) after which a changelog is started.
The thing to be noted here is that for a time interval
(time slice) multiple changes for an inode are recorded only
once (ie. say for 100+ writes on an inode that happens within
the time slice has only a single corresponding entry in the
changelog file). That way we do not bloat up the changelog
and also save lots of writes.
Changelog Format
-----------------
TYPE-I and TYPE-II changes have the gfid on the entity on
which the operation happened. TYPE-III being a entry op
requires the parent gfid and the basename. Changelog format
has been kept to a minimal and it's upto the consumers to
do the heavy loading of figuring out deletes, renames etc..
A single changelog file records all three types of changes,
with each change starting with an identifier ("D": DATA,
"M": METADATA and "E": ENTRY). Option is provided for the
encoding type (See TUNABLES).
Consumers
----------
The only consumer as of today would be geo-replication, although
backup utilities, self-heal, bit-rot detection could be possible
consumers in the future.
CLI
----
By default, change-logging is disabled (the translator is present
in the server graph but does nothing). When enabled (via cli) each
brick starts to log the changes. There are a set of tunable that
can be used to change the translators behaviour:
- enable/disable changelog (disabled by default)
gluster volume set <volume> changelog {on|off}
- set the logging directory (<brick>/.glusterfs/changelogs is the
default)
gluster volume set <volume> changelog-dir /path/to/dir
- select encoding type (binary (default) or ascii)
gluster volume set <volume> encoding {binary|ascii}
- change the rollover time for the logs (60 secs by default)
gluster volume set <volume> rollover-time <secs>
- when secs > 0, changelog file is not open()'d with O_SYNC flag
- and fsync is trigerred periodically every <secs> seconds.
gluster volume set <volume> fsync-interval <secs>
features/changelog: changelog consumer library (libgfchangelog)
A shared library is provided for the consumer of the changelogs
for easy acess via APIs. Application can link against this library
and request for changelog updates. Conversion of binary logs to
human-readable ascii format is also taken care by the library which
keeps a copy of the changelog in application provided working
directory.
Change-Id: I75575fb7f1c53d2bec3dba1a329ea7bb3c628497
BUG: 847839
Original Author: Venky Shankar <vshankar@redhat.com>
Signed-off-by: Avra Sengupta <asengupt@redhat.com>
Reviewed-on: http://review.gluster.org/5127
Tested-by: Gluster Build System <jenkins@build.gluster.com>
Reviewed-by: Vijay Bellur <vbellur@redhat.com>
|
