| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Problem: The CTR xlator records file meta (heat/hardlinks)
into the data. This works fine for files which are created
after ctr xlator is switched ON. But for files which were
created before CTR xlator is ON, CTR xlator is not able to
record either of the meta i.e heat or hardlinks. Thus making
those files immune to promotions/demotions.
Solution: The solution that is implemented in this patch is
do ctr-db heal of all those pre-existent files, using named lookup.
For this purpose we use the inode-xlator context variable option
in gluster.
The inode-xlator context variable for ctr xlator will have the
following,
a. A Lock for the context variable
b. A hardlink list: This list represents the successful looked
up hardlinks.
These are the scenarios when the hardlink list is updated:
1) Named-Lookup: Whenever a named lookup happens on a file, in the
wind path we copy all required hardlink and inode information to
ctr_db_record structure, which resides in the frame->local variable.
We dont update the database in wind. During the unwind, we read the
information from the ctr_db_record and ,
Check if the inode context variable is created, if not we create it.
Check if the hard link is there in the hardlink list.
If its not there we add it to the list and send a update to the
database using libgfdb.
Please note: The database transaction can fail(and we ignore) as there
already might be a record in the db. This update to the db is to heal
if its not there.
If its there in the list we ignore it.
2) Inode Forget: Whenever an inode forget hits we clear the hardlink list in
the inode context variable and delete the inode context variable.
Please note: An inode forget may happen for two reason,
a. when the inode is delete.
b. the in-memory inode is evicted from the inode table due to cache limits.
3) create: whenever a create happens we create the inode context variable and
add the hardlink. The database updation is done as usual by ctr.
4) link: whenever a hardlink is created for the inode, we create the inode context
variable, if not present, and add the hardlink to the list.
5) unlink: whenever a unlink happens we delete the hardlink from the list.
6) mknod: same as create.
7) rename: whenever a rename happens we update the hardlink in list. if the hardlink
was not present for updation, we add the hardlink to the list.
What is pending:
1) This solution will only work for named lookups.
2) We dont track afr-self-heal/dht-rebalancer traffic for healing.
> http://review.gluster.org/#/c/10370/
> Cherry picked from commit cb11dd91a6cc296e4a3808364077f4eacb810e48
> Change-Id: Ia4bbaf84128ad6ce8c3ddd70bcfa82894c79585f
> BUG: 1212037
> Signed-off-by: Joseph Fernandes <josferna@redhat.com>
> Signed-off-by: Dan Lambright <dlambrig@redhat.com>
> Reviewed-on: http://review.gluster.org/10370
> Reviewed-by: Jeff Darcy <jdarcy@redhat.com>
> Tested-by: Gluster Build System <jenkins@build.gluster.com>
> Tested-by: NetBSD Build System
> Reviewed-by: Vijay Bellur <vbellur@redhat.com>
Change-Id: I367aa46c3f4b8f912248fb8be75866507f2538df
BUG: 1219075
Signed-off-by: Joseph Fernandes <josferna@redhat.com>
Signed-off-by: Dan Lambright <dlambrig@redhat.com>
Reviewed-on: http://review.gluster.org/10370
Reviewed-by: Jeff Darcy <jdarcy@redhat.com>
Reviewed-by: Vijay Bellur <vbellur@redhat.com>
Signed-off-by: Joseph Fernandes <josferna@redhat.com>
Reviewed-on: http://review.gluster.org/10615
Tested-by: Vijay Bellur <vbellur@redhat.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
1) ctr_link_consistency option for ctr xaltor is provided so that
the user can choose to switch it on or off.
/* For link consistency we do a double update i.e mark the link
* during the wind and during the unwind we update/delete the link.
* This has a performance hit. We give a choice here whether we need
* link consistency to be spoton or not using link_consistency flag.
* This will have only one link update */
2) In delete the wind time recording is moved to unwind path.
/* Special performance case:
* Updating wind time in unwind for delete. This is done here
* as in the wind path we will not know whether its the last
* link or not. For a last link there is not use to update any
* wind or unwind time!*/
> http://review.gluster.org/#/c/10170/
> Cherry picked from commit 606d9734543208542afcf9df982bf2d560235ef6
> Change-Id: I209472fb816f939db4a868b97ba053b028f17ea6
> BUG: 1217786
> Signed-off-by: Joseph Fernandes <josferna@redhat.com>
> Reviewed-on: http://review.gluster.org/10170
> Reviewed-by: Dan Lambright <dlambrig@redhat.com>
> Tested-by: Gluster Build System <jenkins@build.gluster.com>
> Signed-off-by: Joseph Fernandes <josferna@redhat.com>
Change-Id: I4a89ef80875f36cff91520f712e1f47fde258a63
BUG: 1219066
Signed-off-by: Joseph Fernandes <josferna@redhat.com>
Reviewed-on: http://review.gluster.org/10170
Reviewed-by: Dan Lambright <dlambrig@redhat.com>
Tested-by: Gluster Build System <jenkins@build.gluster.com>
Signed-off-by: Joseph Fernandes <josferna@redhat.com>
Reviewed-on: http://review.gluster.org/10614
Tested-by: NetBSD Build System
Reviewed-by: Vijay Bellur <vbellur@redhat.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This fix will solve the heating of the files during the promotion
or demotion.
Promotion:
~~~~~~~~~
When a file gets promoted it get the current time stamp
during creation only, but following writes or reads during the
migration wont heat the file.
Demotion:
~~~~~~~~
When a file gets demoted it get the wind/unwind time stamp is set to
zero. The following writes or reads during the migration wont heat
the file.
What is remaining ?
~~~~~~~~~~~~~~~~~
Bug 1209129 ( https://bugzilla.redhat.com/show_bug.cgi?id=1209129 )
Inspite of this fix there is still a issue remaining, i.e the heat of
the file is not keep intact during a internal rebalance activity i.e
a rebalance within a tier.
Change-Id: I01e82dc226355599732d40e699062cee7960b0a5
BUG: 1207867
Signed-off-by: Joseph Fernandes <josferna@redhat.com>
Signed-off-by: Dan Lambright <dlambrig@redhat.com>
Signed-off-by: Joseph Fernandes <josferna@redhat.com>
Reviewed-on: http://review.gluster.org/10080
Reviewed-by: Shyamsundar Ranganathan <srangana@redhat.com>
Tested-by: Gluster Build System <jenkins@build.gluster.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
glusterfs relies on Linux uuid implementation, which
API is incompatible with most other systems's uuid. As
a result, libglusterfs has to embed contrib/uuid,
which is the Linux implementation, on non Linux systems.
This implementation is incompatible with systtem's
built in, but the symbols have the same names.
Usually this is not a problem because when we link
with -lglusterfs, libc's symbols are trumped. However
there is a problem when a program not linked with
-lglusterfs will dlopen() glusterfs component. In
such a case, libc's uuid implementation is already
loaded in the calling program, and it will be used
instead of libglusterfs's implementation, causing
crashes.
A possible workaround is to use pre-load libglusterfs
in the calling program (using LD_PRELOAD on NetBSD for
instance), but such a mechanism is not portable, nor
is it flexible. A much better approach is to rename
libglusterfs's uuid_* functions to gf_uuid_* to avoid
any possible conflict. This is what this change attempts.
BUG: 1206587
Change-Id: I9ccd3e13afed1c7fc18508e92c7beb0f5d49f31a
Signed-off-by: Emmanuel Dreyfus <manu@netbsd.org>
Reviewed-on: http://review.gluster.org/10017
Tested-by: Gluster Build System <jenkins@build.gluster.com>
Reviewed-by: Niels de Vos <ndevos@redhat.com>
|
|
|
*************************************************************************
Libgfdb |
*************************************************************************
Libgfdb provides abstract mechanism to record extra/rich metadata
required for data maintenance, such as data tiering/classification.
It provides consumer with API for recording and querying, keeping
the consumer abstracted from the data store used beneath for storing data.
It works in a plug-and-play model, where data stores can be plugged-in.
Presently we have plugin for Sqlite3. In the future will provide recording
and querying performance optimizer. In the current implementation the schema
of metadata is fixed.
Schema:
~~~~~~
GF_FILE_TB Table:
~~~~~~~~~~~~~~~~~
This table has one entry per file inode. It holds the metadata required to
make decisions in data maintenance.
GF_ID (Primary key) : File GFID (Universal Unique IDentifier in the namespace)
W_SEC, W_MSEC : Write wind time in sec & micro-sec
UW_SEC, UW_MSEC : Write un-wind time in sec & micro-sec
W_READ_SEC, W_READ_MSEC : Read wind time in sec & micro-sec
UW_READ_SEC, UW_READ_MSEC : Read un-wind time in sec & micro-sec
WRITE_FREQ_CNTR INTEGER : Write Frequency Counter
READ_FREQ_CNTR INTEGER : Read Frequency Counter
GF_FLINK_TABLE:
~~~~~~~~~~~~~~
This table has all the hardlinks to a file inode.
GF_ID : File GFID (Composite Primary Key)``|
GF_PID : Parent Directory GFID (Composite Primary Key) |-> Primary Key
FNAME : File Base Name (Composite Primary Key)__|
FPATH : File Full Path (Its redundant for now, this will go)
W_DEL_FLAG : This Flag is used for crash consistancy, when a link is unlinked.
i.e Set to 1 during unlink wind and during unwind this record
is deleted
LINK_UPDATE : This Flag is used when a link is changed i.e rename.
Set to 1 when rename wind and set to 0 in rename unwind
Libgfdb API:
~~~~~~~~~~~
Refer libglusterfs/src/gfdb/gfdb_data_store.h
Change-Id: I2e9fbab3878ce630a7f41221ef61017dc43db11f
BUG: 1194753
Signed-off-by: Joseph Fernandes <josferna@redhat.com>
Signed-off-by: Dan Lambright <dlambrig@redhat.com>
Signed-off-by: Joseph Fernandes <josferna@redhat.com>
Reviewed-on: http://review.gluster.org/9683
Tested-by: Gluster Build System <jenkins@build.gluster.com>
Reviewed-by: Vijay Bellur <vbellur@redhat.com>
|
