IF YOU WOULD LIKE TO GET AN ACCOUNT, please write an
email to Administrator. User accounts are meant only to access repo
and report issues and/or generate pull requests.
This is a purpose-specific Git hosting for
BaseALT
projects. Thank you for your understanding!
Только зарегистрированные пользователи имеют доступ к сервису!
Для получения аккаунта, обратитесь к администратору.
Problem:
Currenly the CLI rebalance status command output does not indicate the
'type' of rebalance, i.e. whether a full rebalance or only a fix-layout
was carried out.
Fix: After the rebalance status of all peers is received by the
originator glusterd, alter it to reflect the type of rebalance
before passing it on to the CLI process.
Change-Id: I1940ffda0d36e25e5b33c84a0ea210394cc9e1d3
BUG: 1004744
Signed-off-by: Ravishankar N <ravishankar@redhat.com>
Reviewed-on: http://review.gluster.org/5826
Reviewed-by: Krishnan Parthasarathi <kparthas@redhat.com>
Tested-by: Gluster Build System <jenkins@build.gluster.com>
Reviewed-by: Anand Avati <avati@redhat.com>
Add aggregate status information in <aggregate> section of gluster volume
'rebalance status' and 'remove-brick status cli xml output.
The aggregate status determined based on the most critical level
and the aggregate status will have 'Complete' only when all
individual status are 'Complete'.
Change-Id: Ie805b9dd52fd82fd277c3da9ee91cc8b6dea8212
BUG: 1006813
Signed-off-by: Timothy Asir <tjeyasin@redhat.com>
Reviewed-on: http://review.gluster.org/4950
Tested-by: Gluster Build System <jenkins@build.gluster.com>
Reviewed-by: Kaushal M <kaushal@redhat.com>
The rebalance status was being reset to 'Not started' when add-brick was
performed. This would lead to odd cases where a 'rebalance status' on a
volume would show status as 'not started' but would also include the
rebalance statistics. This also affected the showing of asynchronus task
status in 'volume status' command.
By not resetting the status prevent the above issues from happening.
Since we use the running/not-running of the rebalance process as the
check when performing other operations we can safely leave the rebalance
stats collected on an add-brick.
Change-Id: I4c69d9c789d081c6de7e7a81dd0d4eba2e83ec17
BUG: 1006247
Signed-off-by: Kaushal M <kaushal@redhat.com>
Reviewed-on: http://review.gluster.org/5895
Reviewed-by: Vijay Bellur <vbellur@redhat.com>
Tested-by: Gluster Build System <jenkins@build.gluster.com>
AM_PATH_XML2 needs check for its existence using
aclocal macros - to avoid problems like below when
libxml2 development libraries are not installed
--------------------------------------------------
... GlusterFS autogen ...
Generate gf-error-codes.h ...
`gf-error-codes.h' -> `libglusterfs/src/gf-error-codes.h'
Running aclocal...
configure.ac:524: warning: macro `AM_PATH_XML2' not found in library
Running autoheader...
Running libtoolize...
Running autoconf...
configure.ac:524: error: possibly undefined macro: AM_PATH_XML2
If this token and others are legitimate, please use m4_pattern_allow.
See the Autoconf documentation.
Running automake...
--------------------------------------------------
Change-Id: Ife463c34c45babc1c4c0ed6e8128b5c43419b9b9
BUG: 947226
Signed-off-by: Harshavardhana <harsha@harshavardhana.net>
Reviewed-on: http://review.gluster.org/5962
Reviewed-by: Kaushal M <kaushal@redhat.com>
Tested-by: Gluster Build System <jenkins@build.gluster.com>
Reviewed-by: Vijay Bellur <vbellur@redhat.com>
This reverts commit 215fea41a96479312a5ab8783c13b30ab9fe00fa
Realized soon after merging, that this patch is actually useless.
Checking for inode utilization on dst relative to src is of no use
because dst is already storing linkfile and inode is already consumed
no matter what. We might as well perform the rebalance and save on an
inode on the src node.
Change-Id: I7b8b8d35d9063a710e1bee1c7c6c7739fcc45ad4
Reviewed-on: http://review.gluster.org/5960
Reviewed-by: Harshavardhana <harsha@harshavardhana.net>
Tested-by: Anand Avati <avati@redhat.com>
Currently during `MIGRATE_DATA` we never verified about the total
inode usage among new and old bricks. Such checks are available for
disk space usage but it is also needed for inodes during data
migration. Such a check leads to uniform outcome of file distribution
upon rebalance.
Patch provides:
- Check dst_inodes < src_inodes, a friendly `warning` message to
indicate we have to ignore such an attempt
- Rename __dht_check_free_space() --> __dht_check_free_space_and_inodes()
Change-Id: I7bc4dd8b507883f0fb836300e99f0bb083493f5f
BUG: 982919
Signed-off-by: Harshavardhana <harsha@harshavardhana.net>
Reviewed-on: http://review.gluster.org/5948
Tested-by: Gluster Build System <jenkins@build.gluster.com>
Reviewed-by: Anand Avati <avati@redhat.com>
If kernel supports FUSE_AUTO_INVAL_DATA then it is safe(r) to turn on
--fopen-keep-cache mode by default. Users report significant improvement
in perf by enabling the mode.
Change-Id: Icf9df4b7b43950d7e25302d9c2a1a7d14571a9a9
BUG: 990744
Signed-off-by: Anand Avati <avati@redhat.com>
Reviewed-on: http://review.gluster.org/5770
Tested-by: Gluster Build System <jenkins@build.gluster.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
For rpc requests having large aux group list, allocate large list
on demand. Else use small static array by default.
Without this patch, glusterfsd allocates 140+MB of resident memory
just to get started and initialized.
Change-Id: I3a07212b0076079cff67cdde18926e8f3b196258
Signed-off-by: Anand Avati <avati@redhat.com>
BUG: 953694
Reviewed-on: http://review.gluster.org/5927
Tested-by: Gluster Build System <jenkins@build.gluster.com>
Reviewed-by: Amar Tumballi <amarts@redhat.com>
At the very least, we should PARENT_DOWN on the currently active
graph and disconnect ourselves from glusterd.
Further cleanups underway.
Change-Id: I9276686a84b0975b5ce272b4cbec1b80920d5c5c
BUG: 1004519
Signed-off-by: Anand Avati <avati@redhat.com>
Reviewed-on: http://review.gluster.org/5788
Tested-by: Gluster Build System <jenkins@build.gluster.com>
Reviewed-by: Raghavendra Bhat <raghavendra@redhat.com>
Reviewed-by: Amar Tumballi <amarts@redhat.com>
There is little value in using synctask wrappers for async IO
requests, as STACK_WIND is asynchronous by nature already.
Skip going through synctask for read/write async calls.
Change-Id: Ifde331d7c97e0f33426da6ef4377c5ba70dddc06
BUG: 1009134
Signed-off-by: Anand Avati <avati@redhat.com>
Reviewed-on: http://review.gluster.org/5897
Support the readdirp fop in qemu-block to ensure that image files
are handled correctly when readdirp is enabled. E.g., without
readdirp support, incorrect stat data for formatted files can be
reported back (and cached) via the client.
BUG: 986775
Change-Id: Ibc4bd0b42548953ebe30832f3d853bb68095f0ac
Signed-off-by: Brian Foster <bfoster@redhat.com>
Reviewed-on: http://review.gluster.org/5946
Tested-by: Gluster Build System <jenkins@build.gluster.com>
Reviewed-by: Anand Avati <avati@redhat.com>
With the introduction of inode_ctx_set in fuse as part of 2991503d014,
forget cbk gets called for fuse xlator. Though nothing needs to be done
inf forget_cbk, excessive log messages of the following kind are observed:
[2013-09-16 06:09:50.758063] W [defaults.c:1331:default_forget]
(-->/usr/local/lib/glusterfs/3git/xlator/mount/fuse.so(+0xa1f2) [0x7f51432781f2]
(-->/usr/local/lib/libglusterfs.so.0(inode_unref+0x3c) [0x7f5144e5
816c] (-->/usr/local/lib/libglusterfs.so.0(+0x2d061) [0x7f5144e58061]))) 0-fuse:
xlator does not implement forget_cbk
This patch prevents such log messages from being seen.
Signed-off-by: Vijay Bellur <vbellur@redhat.com>
BUG: 979910
Change-Id: Ie5874138f46822b10ff4213bd1134d78330ec460
Reviewed-on: http://review.gluster.org/5932
Reviewed-by: Anand Avati <avati@redhat.com>
Tested-by: Anand Avati <avati@redhat.com>
This patch introduces task parameters for the asynchronus task shown in
volume status. The parameters are only given for xml output. The
parameters shown currently are,
- source and destination bricks for replace-brick tasks
......
<tasks>
<task>
<type>Replace brick</type>
<id>3d1a1005-9d2e-4ae0-bd62-577bc1d333a3</id>
<status>1</status>
<params>
<srcBrick>archm:/export/test4</srcBrick>
<dstBrick>archm:/export/test-replace1</dstBrick>
</params>
</task>
</tasks>
......
- list of bricks being removed for remove-brick tasks
......
<tasks>
<task>
<type>Remove brick</type>
<id>901c20ca-0da2-41de-8669-5f0caca6b846</id>
<status>1</status>
<params>
<brick>archm:/export/test2</brick>
<brick>archm:/export/test3</brick>
</params>
</task>
</tasks>
......
The changes for non-xml output will be done in a subsequent patch.
Change-Id: I322afe2f83ed8adeddb99f7962c25911204dc204
BUG: 916577
Signed-off-by: Kaushal M <kaushal@redhat.com>
Reviewed-on: http://review.gluster.org/5771
Reviewed-by: Vijay Bellur <vbellur@redhat.com>
Tested-by: Vijay Bellur <vbellur@redhat.com>
Building RPMs from a 'make dist' tarball fails when qemu-block is
enabled. Enabling is done automatically when the glib2 development files
are available (enabled by ./configure).
Manual building with:
$ ./autogen.sh && ./configure && make dist && rpmbuild -ta *.tar.gz
Building in mock works fine, glib2-devel is not installed by default so
the qemu-block xlator gets disabled. This change also adds glib2-devel
to the BuildRequires in the glusterfs.spec file, causing the qemu-block
xlator to be built by default, and included in the glusterfs RPM.
Change-Id: Ibb73628772586d9e07bbfde7a8ff2fc973489086
BUG: 986775
Signed-off-by: Niels de Vos <ndevos@redhat.com>
Reviewed-on: http://review.gluster.org/5896
Tested-by: Gluster Build System <jenkins@build.gluster.com>
Reviewed-by: Anand Avati <avati@redhat.com>
When running multiple mock builds at the same time, the space
requirements under /var/tmp increases as well. Not all systems (like
build.gluster.org) have enough free space in /var/tmp to host two
mock-roots, which causes building to fail randomly.
Change-Id: Iff1064d01893fb35ef59650d4373105068b5d3f7
BUG: 1006269
Signed-off-by: Niels de Vos <ndevos@redhat.com>
Reviewed-on: http://review.gluster.org/5899
Tested-by: Gluster Build System <jenkins@build.gluster.com>
Reviewed-by: Kaleb KEITHLEY <kkeithle@redhat.com>
Reviewed-by: Anand Avati <avati@redhat.com>
New xml tag statusStr added to following gluster cli commands
gluster volume status all --xml (For Task status)
gluster volume rebalance <VOLNAME> status --xml
gluster volume remove-brick <VOLNAME> <BRICK1..> status --xml
Example(volume status all):
<task>
<type>Rebalance</type>
<id>82d8d122-8738-4144-8507-d93fc98b61df</id>
<status>3</status>
<statusStr>completed</statusStr>
</task>
Example(volume rebalance <VOL> status)
<node>
<nodeName>localhost</nodeName>
<files>0</files>
<size>0</size>
<lookups>0</lookups>
<failures>0</failures>
<status>3</status>
<statusStr>completed</statusStr>
</node>
Also modified task status as string instead of showing number
in gluster volume status all
Example:
Status of volume: gv1
Gluster process Port Online Pid
------------------------------------------------------------------------------
Brick sumne.sumne:/gfs/b1 49154 Y 15489
Brick sumne.sumne:/gfs/b2 49155 Y 15493
NFS Server on localhost N/A N 15913
Task ID Status
---- -- ------
Rebalance 82d8d122-8738-4144-8507-d93fc98b61df completed
BUG: 1003521
Change-Id: Ib283016af4c18132fb13fb33d44075782d77823c
Signed-off-by: Aravinda VK <avishwan@redhat.com>
Reviewed-on: http://review.gluster.org/5739
Reviewed-by: Kaushal M <kaushal@redhat.com>
Tested-by: Gluster Build System <jenkins@build.gluster.com>
Problem:
When requests are received on a connection before setvolume
is done, creating frame from the requests fail because there is
no association of the transport with the conn(i.e. xl_private).
xl_private is set only on set_volume. In such cases error response
is not sent from server xlator to that request because of which
operations on mount point are hanging.
Fix:
Set actor return value to RPCSVC_ACTOR_ERROR so that response is sent
even in these cases.
Change-Id: I74d7bc6849fde6c734008d67c1f4bc9d9f7a84f9
BUG: 1006367
Signed-off-by: Pranith Kumar K <pkarampu@redhat.com>
Reviewed-on: http://review.gluster.org/5892
Tested-by: Gluster Build System <jenkins@build.gluster.com>
Reviewed-by: Anand Avati <avati@redhat.com>
CLI now only outputs one XML document for 'status all' only containing
those volumes which are started.
BUG: 1004218
Change-Id: Id4130fe59b3b74475d8bd1cc8134ac59a28f1b7e
Signed-off-by: Kaushal M <kaushal@redhat.com>
Reviewed-on: http://review.gluster.org/5773
Reviewed-by: Vijay Bellur <vbellur@redhat.com>
Tested-by: Gluster Build System <jenkins@build.gluster.com>
save about six minutes per regression by building in parallel
Change-Id: Iaf6fa545e20b0322b10e9e4ab7cc3d4254a0ece5
BUG: 1006269
Signed-off-by: Kaleb S. KEITHLEY <kkeithle@redhat.com>
Reviewed-on: http://review.gluster.org/5845
Tested-by: Gluster Build System <jenkins@build.gluster.com>
Reviewed-by: Niels de Vos <ndevos@redhat.com>
Earlier, a peer running a higher op-version couldn't be probed into a
cluster running at a lower op-version. This created issues when trying
to expand an upgraded cluster. This patch changes this behaviour.
The cluster no longer rejects a peer being probed if its op-version is
higher than the cluster op-version. The peer will reduce its op-version
if it doesn't have any volumes. If the peer contains volumes and needs
to reduce its op-version, it fails the handshake and the probe fails.
Change-Id: I12c6c873922799e1557b7184e956baea643d0dea
BUG: 1005038
Signed-off-by: Kaushal M <kaushal@redhat.com>
Reviewed-on: http://review.gluster.org/5715
Tested-by: Gluster Build System <jenkins@build.gluster.com>
Reviewed-by: Vijay Bellur <vbellur@redhat.com>
`use-readdirp` has an optional argument in argp -
specifying just `--use-readdirp` command line should not
'turn off' readdirp, since that undermines the meaning
of such an argument.
Change-Id: I965d87e29bd0d61997d9be96fa698e270a2ee173
BUG: 983477
Signed-off-by: Harshavardhana <harsha@harshavardhana.net>
Reviewed-on: http://review.gluster.org/5851
Tested-by: Gluster Build System <jenkins@build.gluster.com>
Reviewed-by: Anand Avati <avati@redhat.com>
The current self-healing algorithm is ignoring missing directories
for assigning new layout. When lookup() is racing against mkdir()
or when self-healing a half-done mkdir(), the layout assignment split
must happen based on the final number of directories, and not the
currently existing number of directories (because we finish mkdir()
of missing directories before hash layout assignment).
Without this fix, concurrent mkdir() and lookup() will step on
each others feet, create a messed up layout on disk, and end up
with different in-memory layouts.
Once two clients have different in-memory layouts, creation of
subdirectory will not arbitrate on the same hashed subvolume and will
result in GFID mismatch of the sub-directory.
Change-Id: Ia47acad67c265060405984c822b4d37512b9dbb3
BUG: 907072
Signed-off-by: Anand Avati <avati@redhat.com>
Reviewed-on: http://review.gluster.org/5849
Tested-by: Gluster Build System <jenkins@build.gluster.com>
Reviewed-by: Amar Tumballi <amarts@redhat.com>
Reviewed-by: Peter Portante <pportant@redhat.com>
Tested-by: Peter Portante <pportant@redhat.com>
Commit a482e422 added glusterd.info to the %files section of the
glusterfs-server package. However, this file was listed already. This
causes the following warning when building the RPMs:
warning: File listed twice: /var/lib/glusterd/glusterd.info
Merging the attributes for /var/lib/glusterd/glusterd.info into one line
prevents this warning.
Change-Id: I0d518ec186a8725dc4c5cba00b60da83fdadf103
BUG: 1005161
Signed-off-by: Niels de Vos <ndevos@redhat.com>
Reviewed-on: http://review.gluster.org/5836
Reviewed-by: Kaleb KEITHLEY <kkeithle@redhat.com>
Tested-by: Gluster Build System <jenkins@build.gluster.com>
Reviewed-by: Anand Avati <avati@redhat.com>
The volfile parser thread safe by guarding the parsing phase
in a mutex. Thread safety becomes a problem when there are multiple
glfs_t objects created by gfapi and all of them potentially parse
the respective volfiles at the same time.
Change-Id: I4376019c4956994b72397ab36e6ac3ce849797ec
BUG: 1004519
Signed-off-by: Anand Avati <avati@redhat.com>
Reviewed-on: http://review.gluster.org/5790
Reviewed-by: Raghavendra G <rgowdapp@redhat.com>
Reviewed-by: Raghavendra Bhat <raghavendra@redhat.com>
Tested-by: Gluster Build System <jenkins@build.gluster.com>
Reviewed-by: Anand Avati <avati@gluster.org>
The flags passed on to open and create calls were not being saved
in the fd_t object, hence the fd migration was failing.
Change-Id: I486bb818477fe4c393d64a711534a082162a0e53
BUG: 1005159
Signed-off-by: Poornima <pgurusid@redhat.com>
Reviewed-on: http://review.gluster.org/5837
Reviewed-by: Amar Tumballi <amarts@redhat.com>
Reviewed-by: Raghavendra Bhat <raghavendra@redhat.com>
Tested-by: Gluster Build System <jenkins@build.gluster.com>
Reviewed-by: Anand Avati <avati@redhat.com>
config service availability for clients.
Backupvolfile server as it stands is slow and prone to errors
with mount script and its combination with RRDNS. Instead in
theory it should use all the available nodes in 'trusted pool'
by default (Right now we don't have a mechanism in place for
this)
Nevertheless this patch provides a scenario where a list of
volfile-server can be provided on command as shown below
-----------------------------------------------------------------
$ glusterfs -s server1 .. -s serverN --volfile-id=<volname> \
<mount_point>
-----------------------------------------------------------------
OR
-----------------------------------------------------------------
$ mount -t glusterfs -obackup-volfile-servers=<server2>: \
<server3>:...:<serverN> <server1>:/<volname> <mount_point>
-----------------------------------------------------------------
Here ':' is used as a separator for mount script parsing
Now these will be remembered and recursively attempted for
fetching vol-file until exhausted. This would ensure that the
clients get 'volume' configs in a consistent manner avoiding the
need to poll through RRDNS.
Change-Id: If808bb8a52e6034c61574cdae3ac4e7e83513a40
BUG: 986429
Signed-off-by: Harshavardhana <harsha@harshavardhana.net>
Reviewed-on: http://review.gluster.org/5400
Tested-by: Gluster Build System <jenkins@build.gluster.com>
Reviewed-by: Anand Avati <avati@redhat.com>
This reverts commit 91e8103794349ebdd578b77987a089456a59c52b
Reverting as regression tests are failing after this merge.
Change-Id: I9dcd430b2f9a431c3c01deab1a205aed297216a3
Reviewed-on: http://review.gluster.org/5842
Reviewed-by: Vijay Bellur <vbellur@redhat.com>
Tested-by: Vijay Bellur <vbellur@redhat.com>
save save about five minutes per regression
Change-Id: I75d91cfef63347ca756b9e1fedcb697e9537edc0
Signed-off-by: Kaleb S. KEITHLEY <kkeithle@redhat.com>
Reviewed-on: http://review.gluster.org/5828
Tested-by: Gluster Build System <jenkins@build.gluster.com>
Reviewed-by: Anand Avati <avati@redhat.com>
In case of nested symlink resolution, implement an upper
bound on the number of such nested levels the resolver will
descend. This limit is arbitrary, and set to 2048 nested
levels.
Change-Id: I264e5bd60d317eda97f4e6f49bd5d8694f8664a9
BUG: 1004100
Signed-off-by: Anand Avati <avati@redhat.com>
Reviewed-on: http://review.gluster.org/5768
Tested-by: Gluster Build System <jenkins@build.gluster.com>
Reviewed-by: Raghavendra Talur <rtalur@redhat.com>
Reviewed-by: Vijay Bellur <vbellur@redhat.com>
The callback structures in both protocol/client and glusterfsd,
gfapi used the same name for the actor table - gluster_cbk_actors.
CBKs are required only for the management connection, and the
actors of protocol/client are NOP functions. This supposed-to-be
NOP function dispatch tabble is actually ending up pointing to
the actor table of glusterfsd or gfapi.
These functions, even though set wrongly, are not even expected
to be called through the protocol/client callback path. Glusterd
however sends the FETCHSPEC (and other) notify callbacks to *all*
connected clients unconditionally, and there is a small period
of time when protocol/client is connected to glusterd for
PORTMAP query. If the FETCHSPEC callback notify is issued in
this window of time, we end up calling the wrong actor in the
client side resulting in a crash.
Change-Id: I605ff7df64c7faf4607369bbf275aedec28e1778
BUG: 1004091
Signed-off-by: Anand Avati <avati@redhat.com>
Reviewed-on: http://review.gluster.org/5767
Tested-by: Gluster Build System <jenkins@build.gluster.com>
Reviewed-by: Amar Tumballi <amarts@redhat.com>
With entry retries MAX_RETRIES is the threashold for the number
of times we retry a changelog. After hitting this limit, the
changelog is marked as processed after warning for incompleteness
in the log.
Change-Id: I777b71b4cdaa63335d9c936f2d64e5c0264a0f26
BUG: 1003803
Signed-off-by: Venky Shankar <vshankar@redhat.com>
Reviewed-by: Amar Tumballi <amarts@redhat.com>
Tested-by: Amar Tumballi <amarts@redhat.com>
Reviewed-on: http://review.gluster.org/5758
Tested-by: Gluster Build System <jenkins@build.gluster.com>
Reviewed-by: Avra Sengupta <asengupt@redhat.com>
Tested-by: Avra Sengupta <asengupt@redhat.com>
Reviewed-by: Anand Avati <avati@redhat.com>
This is certainly a "hack" for cases when after a rename
the gfid is unlinked (which is now pointed by a different
pargfid/basename). This was osbserved when a file edited
(uisng vi/vim) on the glusterfs mount point which has a
geo-replication session running.
Change-Id: I0a5d9f0f549aa4403671fa1f11e95510142e5720
BUG: 1003800
Signed-off-by: Avra Sengupta <asengupt@redhat.com>
Signed-off-by: Venky Shankar <vshankar@redhat.com>
Reviewed-by: Amar Tumballi <amarts@redhat.com>
Tested-by: Amar Tumballi <amarts@redhat.com>
Reviewed-on: http://review.gluster.org/5755
Tested-by: Gluster Build System <jenkins@build.gluster.com>
Reviewed-by: Anand Avati <avati@redhat.com>
entry operation on the slave was using source parent gfid
and basename when renames are overloaded to use regular file
creation. This patch fixes the issue by using the destination
parent gfid and basename for these cases.
Change-Id: I1a4e8df7f07905224ce44ef5abd6f180234285ab
BUG: 1003800
Tested-by: Amar Tumballi <amarts@redhat.com>
Signed-off-by: Venky Shankar <vshankar@redhat.com>
Reviewed-on: http://review.gluster.org/5754
Tested-by: Gluster Build System <jenkins@build.gluster.com>
Reviewed-by: Avra Sengupta <asengupt@redhat.com>
Tested-by: Avra Sengupta <asengupt@redhat.com>
Reviewed-by: Anand Avati <avati@redhat.com>
Slave's xtime is now stored on the master itself (and that too only on
the root), which implies it cannot be propogated to the cascaded slave.
Thus the intermediate master now makes use of it's own volume information
to propogate volume-mark and xtime.
On starting Geo-Replication "geo-replication.ignore-pid-check" marker
option is enabled, which is an override for the client-pid check in
marker. This options triggers marker update only for geo-replication
auxillary mount (client-pid == -1). Since gsyncd not does setxattr()
directly on the bricks, this option won't trigger a chain of spurious
metadata updates that would need to be processed by gsyncd.
Change-Id: If50c5ef275dfb6b4ff4fd35be2565587e2fdf3e1
BUG: 996371
Original Author: Venky Shankar <vshankar@redhat.com>
Signed-off-by: Venky Shankar <vshankar@redhat.com>
Reviewed-on: http://review.gluster.org/5592
Tested-by: Gluster Build System <jenkins@build.gluster.com>
Reviewed-by: Avra Sengupta <asengupt@redhat.com>
Tested-by: Avra Sengupta <asengupt@redhat.com>
Reviewed-by: Anand Avati <avati@redhat.com>
This is required by Geo-Replication that does auxillary mount
with client-pid as -1 (which has special treatment at specific
places in GlusterFS), to trigger xtime updates on the intermediate
master in a cascading setup.
Marker too had a check to "not" mark updates for geo-replication's
auxillary mounts. With the new geo-replication design, xtimes are
not set by the master on the slave for all entities. Due to this
cascading setups were broken.
This patch introduces "geo-replication.ignore-pid-check" option
as a "override" for the client-pid check for gsyncd's client-pid.
When this options is enabled, marker start "marking" even if the
updates are from the special client.
Geo-Replication on the detection of itself being an intermediate
master, enables this option.
Change-Id: I9f7140edd12fef5480595ee0f93f35b94cdb8345
BUG: 996371
Signed-off-by: Venky Shankar <vshankar@redhat.com>
Reviewed-on: http://review.gluster.org/5591
Tested-by: Gluster Build System <jenkins@build.gluster.com>
Reviewed-by: Avra Sengupta <asengupt@redhat.com>
Tested-by: Avra Sengupta <asengupt@redhat.com>
Reviewed-by: Anand Avati <avati@redhat.com>
This mode introduces and optimized recovery of master,
used during failover-failback cases. Files are only
transferred if they are identified as changed on the slave.
A normal failback mechanism would be just to reverse the
direction of geo-replication syncing. Doing this is the
correct way but could be time consuming. "Recovery" mode
optimizes this by identifying files (as skipping them) duing
the one-shot crawl.
NOTE: Use this mode when you are _absolutely_ sure that the
namespace on the master is intact, ie. there is number of files
in master is equal to the number of files in the slave, but the
slave was written to (after a failover) when master was brought
back up.
- gluster volume geo <MV> <SLAVE>::<SV> config special_sync_mode recover
- gluster volume geo <MV> <SLAVE>::<SV> start
Change-Id: Idb698949e4e86698b15d04a35d33095562f51714
BUG: 994462
Original Author: Venky Shankar <vshankar@redhat.com>
Signed-off-by: Venky Shankar <vshankar@redhat.com>
Reviewed-on: http://review.gluster.org/5515
Tested-by: Gluster Build System <jenkins@build.gluster.com>
Reviewed-by: Avra Sengupta <asengupt@redhat.com>
Tested-by: Avra Sengupta <asengupt@redhat.com>
Reviewed-by: Anand Avati <avati@redhat.com>
