glusterfs.git/xlators/features, branch v3.8rc2

features/shard: Get hard-link-count in {unlink,rename}_cbk before deleting shards

2016-05-24T08:57:46+00:00

        Backport of http://review.gluster.org/#/c/14334/

Change-Id: Iff0e90bee22e20c309eaea6c6a19e4fa6e101ed7
BUG: 1337839
Signed-off-by: Krutika Dhananjay 
Reviewed-on: http://review.gluster.org/14451
Smoke: Gluster Build System 
NetBSD-regression: NetBSD Build System 
CentOS-regression: Gluster Build System 
Reviewed-by: Niels de Vos

features/worm: When disabled, worm xlator must directly pass fops to its child xlator

2016-05-24T08:34:45+00:00

Backport of http://review.gluster.org/#/c/14182/

> Change-Id: I56ea46c2743636ef7af5c85469cb2e186eadf21f
> BUG: 1332460
> Signed-off-by: Krutika Dhananjay 
> Reviewed-on: http://review.gluster.org/14182
> Smoke: Gluster Build System 
> Reviewed-by: Joseph Fernandes
> NetBSD-regression: NetBSD Build System 
> CentOS-regression: Gluster Build System 
> Reviewed-by: Niels de Vos 
> Signed-off-by: Joseph Fernandes 

Change-Id: Ic3995e8e81ae22a0c31167e356b203acfff4194d
BUG: 1334836
Signed-off-by: Joseph Fernandes 
Reviewed-on: http://review.gluster.org/14367
Smoke: Gluster Build System 
NetBSD-regression: NetBSD Build System 
CentOS-regression: Gluster Build System 
Reviewed-by: Joseph Fernandes
Reviewed-by: Niels de Vos

index: Fix compiler warning

2016-05-24T08:32:57+00:00

Problem:
----------------------------
index.c:420:18: warning: comparison of unsigned enum expression < 0 is
always false [-Wtautological-compare]
        if (type < 0 || type >= XATTROP_TYPE_END)
            ~~~~ ^ ~
1 warning generated.
----------------------------
Fix:
Add a -1 value in the enumeration and use that in if-statements.

Cherry picked from commit e221827c18ad4161a8e54ad67c9128ad234b8c03:
> BUG: 1335232
> Change-Id: I830a565b8d96d7b51ff775e94aa23dd567ffe0d9
> Signed-off-by: Niels de Vos 
> Reviewed-on: http://review.gluster.org/14409
> Smoke: Gluster Build System 
> Reviewed-by: Ravishankar N 
> CentOS-regression: Gluster Build System 
> NetBSD-regression: NetBSD Build System 
> Reviewed-by: Kaleb KEITHLEY 

The backport prevens the need for commit 137bd83 to be backported.
Commit e221827 undoes the ugly type-casting.

Change-Id: I65911a05de5f8b5634f1e8cf359c38dad7d125aa
BUG: 1336268
Signed-off-by: Ravishankar N 
Signed-off-by: Niels de Vos 
Reviewed-on: http://review.gluster.org/14346
Tested-by: Niels de Vos 
Smoke: Gluster Build System 
NetBSD-regression: NetBSD Build System 
CentOS-regression: Gluster Build System 
Reviewed-by: Kaleb KEITHLEY

leases: Send "this" as cookie to the timer handler

2016-05-20T10:57:21+00:00

Issue:
timer-wheel implementation doesn't set the THIS to point
to the xlator who registered the timer, before calling
the handler. Hence referencing to THIS is any timer handler
will point to the global_xlator.

Fix: This is a bug from the timer wheel, but until that gets
fixed passing "this" as a cookie to the timer handler.

Backport of http://review.gluster.org/#/c/14370/
Change-Id: Ife1be56dc100372f9211e8fc7a885ac717cbcf47
BUG: 1337638
Signed-off-by: Poornima G 
Reviewed-on: http://review.gluster.org/14370
Smoke: Gluster Build System 
NetBSD-regression: NetBSD Build System 
CentOS-regression: Gluster Build System 
Reviewed-by: Raghavendra Talur 
Reviewed-by: Pranith Kumar Karampuri 
(cherry picked from commit 84924ee0ef7cb557c54a9d467364ded86b34f40d)
Reviewed-on: http://review.gluster.org/14431

leases: Fix failure of RD_lease request in certain scenario

2016-05-20T10:57:00+00:00

Issue:
When RW_lease exists for a given lease id and another request for
RD_lease comes in with the same lease id, it gets rejected.

Fix:
Allow the RD_lease when RW_lease is present iff its from the
same lease id.

Backport of http://review.gluster.org/#/c/14374/
Change-Id: Id8ad542c649497f5e58f2acf5681c640d9202046
BUG: 1337638
Signed-off-by: Poornima G 
Reviewed-on: http://review.gluster.org/14374
Smoke: Gluster Build System 
NetBSD-regression: NetBSD Build System 
Reviewed-by: Raghavendra Talur 
CentOS-regression: Gluster Build System 
(cherry picked from commit 005d18a623dc49397e2a83a7fb80607ae0a9aed9)
Reviewed-on: http://review.gluster.org/14430

leases: Do not init the threads when lease is disabled

2016-05-20T09:33:00+00:00

Backport of http://review.gluster.org/#/c/14360/

Change-Id: I08c4caf94bf3dfceba6f7d3cc8945c61d9b12dbc
BUG: 1337638
Signed-off-by: Poornima G 
Reviewed-on: http://review.gluster.org/14360
Smoke: Gluster Build System 
NetBSD-regression: NetBSD Build System 
CentOS-regression: Gluster Build System 
Reviewed-by: Jeff Darcy 
Signed-off-by: Poornima G 
Reviewed-on: http://review.gluster.org/14433
Reviewed-by: Raghavendra Talur

leases: Fix the notify being lost

2016-05-20T09:32:34+00:00

The notify() implemented doesn't really add any value
and the EXIT_IF_LEASES_OFF in notify() can cause the
notifications to be lost if leases if off.

Hence remove the notify definition from leases,
also fix a log level.

Backport of http://review.gluster.org/#/c/14275/

Change-Id: I8bf30d26f08034a16a450ed811dcf8ac57d70a94
BUG: 1337638
Signed-off-by: Poornima G 
Reviewed-on: http://review.gluster.org/14275
Smoke: Gluster Build System 
NetBSD-regression: NetBSD Build System 
Reviewed-by: Raghavendra Talur 
CentOS-regression: Gluster Build System 
Reviewed-by: Raghavendra G 
(cherry picked from commit 58c577994b2bf67b6f8d9749d9c7d678edfc9ceb)
Reviewed-on: http://review.gluster.org/14429

features/bitrot: Introduce scrubber monitor thread

2016-05-20T06:18:25+00:00

The patch does following changes.

 1. Introduce scrubber monitor thread.
 2. Move scrub status related APIs to separate file
    and make part of libbitrot library.

Problem:
     Earlier, each child of the scrubber was maintaining
the state machine and hence there was no way to track
the start and end time of scrubbing as each brick has
it's own start and end time. Also each brick was maintaining
it's own timer wheel instance. It was also not possible
to get scrubbed files count per session as we could not
get last child which finishes scrubbing to reset it to
zero.

Solution:
    Introduce scrubber monitor thread. It does following.

 1. Maintains the scrubber state machine. Earlier each
    child had it's own state machine. Now, only monitor
    maintains on behalf of all it's children.
 2. Maintains the timer wheel instance. Earlier each
    child had it's own timer wheel instance. Now, only
    monitor maintains on behalf of all it's children.

As a result, we can track the scrub statistics easily
and correctly.

Upstream master:
http://review.gluster.org/14044
http://review.gluster.org/#/c/14146

Upstream 3.7:
http://review.gluster.org/14140

NOTE: The patch #14146 is a compilation warning not detected
in master branch and detected only in 3.7 branch. Since the
compilation warning is introduced by patch #14044, the above
two backports are made into this single patch in 3.7 and the
same is backported to 3.8.

BUG: 1332839
Change-Id: I1cec0b229c851fec793a1288c23e43bcec755cba
Signed-off-by: Kotresh HR 
Reviewed-on: http://review.gluster.org/14201
Smoke: Gluster Build System 
NetBSD-regression: NetBSD Build System 
CentOS-regression: Gluster Build System 
Reviewed-by: Venky Shankar

core: assorted typos and spelling mistakes reported by Debian lintian

2016-05-18T13:55:02+00:00

Also missing bang (!) in #!/bin/bash in shell scripts.

Backport of
> http://review.gluster.org/#/c/14398/
> BUG: 1336793
> Change-Id: I567a4be8f0f31f6285550f243fe802895f6bc43b

Change-Id: I003554daede9937a41010e085c74e77957c5c608
BUG: 1336794
Reported-by: Patrick Matthäi 
Signed-off-by: Kaleb S KEITHLEY 
Reviewed-on: http://review.gluster.org/14400
NetBSD-regression: NetBSD Build System 
Smoke: Gluster Build System 
CentOS-regression: Gluster Build System

Revert "features/shard: Make o-direct writes work with sharding"

2016-05-14T14:09:31+00:00

        Backport of: http://review.gluster.org/#/c/14328/

This reverts commit c272c71391cea9db817f4e7e38cfc25a7cff8bd5.

This is for two reasons:
1) It introduces high fop latencies
2) Even with the patch, there is no true odirect behavior since the
   workaround in the patch doesn't reduce the caching done in kernel's
   page cache as far as writes on anon fds associated with individual
   shards is concerned.

Change-Id: I4137816a8bff9f0f77d42041a2d17e63dff82b5d
BUG: 1335822
Signed-off-by: Krutika Dhananjay 
Reviewed-on: http://review.gluster.org/14330
NetBSD-regression: NetBSD Build System 
CentOS-regression: Gluster Build System 
Smoke: Gluster Build System 
Reviewed-by: Pranith Kumar Karampuri