On Tue, Dec 17, 2013 at 03:46:53PM +0800, Robin Dong wrote:
Reviewed-by: Robin Dong san...@taobao.com
Applied
Yuan
--
sheepdog mailing list
sheepdog@lists.wpkg.org
http://lists.wpkg.org/mailman/listinfo/sheepdog
For now we can only call them in the main thread, which is designed for long
running or infrequent events. This would be inefficient if we want to deal with
short running and frequent events that register/unregister the events in the
worker thread
1. avoid to be trapped to main thread for
At Tue, 17 Dec 2013 16:29:43 +0800,
Liu Yuan wrote:
For now we can only call them in the main thread, which is designed for long
running or infrequent events. This would be inefficient if we want to deal
with
short running and frequent events that register/unregister the events in the
On Tue, Dec 17, 2013 at 05:34:44PM +0900, Hitoshi Mitake wrote:
At Tue, 17 Dec 2013 16:29:43 +0800,
Liu Yuan wrote:
For now we can only call them in the main thread, which is designed for long
running or infrequent events. This would be inefficient if we want to deal
with
short
On Tue, Dec 17, 2013 at 04:00:05PM +0900, Hitoshi Mitake wrote:
At Tue, 17 Dec 2013 10:56:00 +0900,
Hitoshi Mitake wrote:
This update backports an implementation of '-o' option in 'vdi track'. It is
useful for determining which nodes have a specified object. When 'vdi check'
fails
On Tue, Dec 17, 2013 at 05:52:36PM +0900, Hitoshi Mitake wrote:
This update contains 4 bugfixes related to memory leak and error code
handling.
Hitoshi Mitake (1):
sheep: let init_vdi_state() return correct value
Liu Yuan (1):
sheep: check EAGAIN for all the sd_lock helpers
Hi sheepdog users and developers,
I released v0.7.6-rc0 of stable branch. You can download a source
archive from these URLs:
tar.gz: https://github.com/sheepdog/sheepdog/archive/v0.7.6-rc0.tar.gz
zip: https://github.com/sheepdog/sheepdog/archive/v0.7.6-rc0.zip
This update contains many
Hi sheepdog users and developers,
I released v0.7.6-rc0 of stable branch. You can download a source
archive from these URLs:
tar.gz: https://github.com/sheepdog/sheepdog/archive/v0.7.6-rc0.tar.gz
zip: https://github.com/sheepdog/sheepdog/archive/v0.7.6-rc0.zip
This update contains many
...
Dec 17 19:04:27 ERROR [net 11088] do_read(220) connection is closed (48 bytes
left)
Dec 17 19:04:27 ERROR [net 11088] rx_work(684) failed to read a header
...
This is quit annoying since nothing wrong with it and sheep.log is flooded with
these messages.
Signed-off-by: Liu Yuan
This function is buggy yet tricky to debug. For now we can't pass following
script:
for i in `seq 0 2`; do
sheep/sheep -D -n -z $i -p $((7000+$i)) -c local store/$i
done
sleep 1
dog/dog cluster format -c 3
qemu-img convert linux-0.2.img sheepdog:test
dog/dog vdi snapshot test -s snap
This function is buggy yet tricky to debug. For now we can't pass following
script:
for i in `seq 0 2`; do
sheep/sheep -D -n -z $i -p $((7000+$i)) -c local store/$i
done
sleep 1
dog/dog cluster format -c 3
qemu-img convert linux-0.2.img sheepdog:test
dog/dog vdi snapshot test -s snap
On Tue, Dec 17, 2013 at 06:06:02PM +0900, Hitoshi Mitake wrote:
Hi sheepdog users and developers,
I released v0.7.6-rc0 of stable branch. You can download a source
archive from these URLs:
tar.gz: https://github.com/sheepdog/sheepdog/archive/v0.7.6-rc0.tar.gz
zip:
On Tue, Dec 17, 2013 at 08:02:02PM +0800, Robin Dong wrote:
The new function sd_inode_write_vid_only will only write 'vid' in 'idx' for
non-hyper-volume vid. For hyper-volume vdi, we write the 'vid' in function
sd_inode_set_vid() only when this 'vid' is exists.
Signed-off-by: Robin Dong
Previous protection scheme of wi-nr_thread in work.c was
unclear. This patch let work.c protect wi-nr_thread by
wi-startup_lock.
Signed-off-by: Hitoshi Mitake mitake.hito...@lab.ntt.co.jp
---
lib/work.c | 5 -
1 file changed, 4 insertions(+), 1 deletion(-)
diff --git a/lib/work.c
Current work queue mechanism of libsheepdog has some little but bother problems
related to coding style, functionality, and performance. This patchset solves
them.
v2:
- preserve correct locking schemes for the data structures
Hitoshi Mitake (4):
lib, sheep: exclude stuff for tracing when it
wi-nr_threads is protected by wi-startup_lock now, we can make the
critical section in queue_work() smaller.
Signed-off-by: Hitoshi Mitake mitake.hito...@lab.ntt.co.jp
---
lib/work.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/lib/work.c b/lib/work.c
index
From: Hitoshi Mitake mitake.hito...@lab.ntt.co.jp
The current build process of sheepdog compiles stuff for tracing even
if tracing is disabled. Basically they are not harmful but causes
memory consumption (tid_map), we should exlucde them when tracing is
disabled.
In addition, this patch adds a
This patch removes confusing names in work.c:
1. before: wi-nr_workers, after: wi-nr_queued_work
2. before: struct worker_info, after: struct wq_info
Also fixes obsolete comments.
Signed-off-by: Hitoshi Mitake mitake.hito...@lab.ntt.co.jp
---
lib/work.c | 40
At Tue, 17 Dec 2013 21:43:14 +0800,
Liu Yuan wrote:
On Tue, Dec 17, 2013 at 06:06:02PM +0900, Hitoshi Mitake wrote:
Hi sheepdog users and developers,
I released v0.7.6-rc0 of stable branch. You can download a source
archive from these URLs:
tar.gz:
At Tue, 17 Dec 2013 21:36:17 +0800,
Liu Yuan wrote:
This function is buggy yet tricky to debug. For now we can't pass following
script:
for i in `seq 0 2`; do
sheep/sheep -D -n -z $i -p $((7000+$i)) -c local store/$i
done
sleep 1
dog/dog cluster format -c 3
qemu-img convert
Dec 17 23:38:30 ERROR [gway 5442] default_read_from_path(287) failed to read
object 7c2b25, path=/tmp/store/0/obj/007c2b25, offset=0,
size=4194304, result=-1, Invalid argument
Dec 17 23:38:30 ERROR [gway 5442] err_to_sderr(115) oid=7c2b25,
Invalid argument
Dec 17
At Tue, 17 Dec 2013 23:47:59 +0800,
Liu Yuan wrote:
Dec 17 23:38:30 ERROR [gway 5442] default_read_from_path(287) failed to read
object 7c2b25, path=/tmp/store/0/obj/007c2b25, offset=0,
size=4194304, result=-1, Invalid argument
Dec 17 23:38:30 ERROR [gway 5442]
On Tue, Dec 17, 2013 at 11:55:16PM +0900, Hitoshi Mitake wrote:
At Tue, 17 Dec 2013 21:36:17 +0800,
Liu Yuan wrote:
This function is buggy yet tricky to debug. For now we can't pass following
script:
for i in `seq 0 2`; do
sheep/sheep -D -n -z $i -p $((7000+$i)) -c local
On Wed, Dec 18, 2013 at 12:54:48AM +0900, Hitoshi Mitake wrote:
At Tue, 17 Dec 2013 23:47:59 +0800,
Liu Yuan wrote:
Dec 17 23:38:30 ERROR [gway 5442] default_read_from_path(287) failed to
read object 7c2b25, path=/tmp/store/0/obj/007c2b25,
offset=0, size=4194304,
On Tue, Dec 17, 2013 at 09:26:50PM +0800, Liu Yuan wrote:
...
Dec 17 19:04:27 ERROR [net 11088] do_read(220) connection is closed (48
bytes left)
Dec 17 19:04:27 ERROR [net 11088] rx_work(684) failed to read a header
...
This is quit annoying since nothing wrong with it and sheep.log is
On Tue, Dec 17, 2013 at 11:42:53PM +0900, Hitoshi Mitake wrote:
From: Hitoshi Mitake mitake.hito...@lab.ntt.co.jp
The current build process of sheepdog compiles stuff for tracing even
if tracing is disabled. Basically they are not harmful but causes
memory consumption (tid_map), we should
On Tue, Dec 17, 2013 at 11:42:54PM +0900, Hitoshi Mitake wrote:
Previous protection scheme of wi-nr_thread in work.c was
unclear. This patch let work.c protect wi-nr_thread by
wi-startup_lock.
I don't understand what you meant by 'unclear'. Could make it clear that
what exactly you think is
At Wed, 18 Dec 2013 11:36:41 +0800,
Liu Yuan wrote:
On Tue, Dec 17, 2013 at 11:42:53PM +0900, Hitoshi Mitake wrote:
From: Hitoshi Mitake mitake.hito...@lab.ntt.co.jp
The current build process of sheepdog compiles stuff for tracing even
if tracing is disabled. Basically they are not
At Wed, 18 Dec 2013 11:41:39 +0800,
Liu Yuan wrote:
On Tue, Dec 17, 2013 at 11:42:54PM +0900, Hitoshi Mitake wrote:
Previous protection scheme of wi-nr_thread in work.c was
unclear. This patch let work.c protect wi-nr_thread by
wi-startup_lock.
I don't understand what you meant by
Current work queue mechanism of libsheepdog has some little but bother problems
related to coding style, functionality, and performance. This patchset solves
them.
v3:
- in the 3rd patch, add a new mutex for protecting nr_threads
v2:
- preserve correct locking schemes for the data structures
The current build process of sheepdog compiles stuff for tracing even
if tracing is disabled. Basically they are not harmful but causes
memory consumption (tid_map), we should exlucde them when tracing is
disabled.
In addition, this patch adds a new mutex tid_map_lock for protecting
tid_map.
From: Liu Yuan namei.u...@gmail.com
Also add Robin and myself to the contributor list of this file
Reviewed-by: Hitoshi Mitake mitake.hito...@lab.ntt.co.jp
Reviewed-by: Robin Dong san...@taobao.com
Signed-off-by: Liu Yuan namei.u...@gmail.com
---
sheep/http/kv.c | 101
From: Hitoshi Mitake mitake.hito...@gmail.com
Previous protection scheme of wi-nr_thread in work.c was
unclear because wi-startup_lock was also used for protecting it
during workqueue grow/shrink. This patch let work.c protect
wi-nr_thread by the new wi-workers_lock.
Signed-off-by: Hitoshi
From: Hitoshi Mitake mitake.hito...@gmail.com
wi-nr_threads is protected by wi-startup_lock now, we can make the
critical section in queue_work() smaller.
Signed-off-by: Hitoshi Mitake mitake.hito...@lab.ntt.co.jp
---
lib/work.c |2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff
From: Hitoshi Mitake mitake.hito...@gmail.com
This patch removes confusing names in work.c:
1. before: wi-nr_workers, after: wi-nr_queued_work
2. before: struct worker_info, after: struct wq_info
Also fixes obsolete comments.
Signed-off-by: Hitoshi Mitake mitake.hito...@lab.ntt.co.jp
---
At Wed, 18 Dec 2013 13:35:21 +0800,
Liu Yuan wrote:
On Wed, Dec 18, 2013 at 01:13:15PM +0900, Hitoshi Mitake wrote:
At Wed, 18 Dec 2013 11:36:41 +0800,
Liu Yuan wrote:
On Tue, Dec 17, 2013 at 11:42:53PM +0900, Hitoshi Mitake wrote:
From: Hitoshi Mitake
On Wed, Dec 18, 2013 at 01:35:21PM +0800, Liu Yuan wrote:
On Wed, Dec 18, 2013 at 01:13:15PM +0900, Hitoshi Mitake wrote:
At Wed, 18 Dec 2013 11:36:41 +0800,
Liu Yuan wrote:
On Tue, Dec 17, 2013 at 11:42:53PM +0900, Hitoshi Mitake wrote:
From: Hitoshi Mitake
This update contains an important bugfix of cluster snaphsot feature. If you are
a user of cluster snapshot, please be aware of it. It will be included in
v0.7.6-rc1.
Liu Yuan (1):
sheep: remove sha1_from_buffer()
dog/farm/sha1_file.c |4 ++--
include/sha1.h |2 +-
lib/sha1.c
From: Liu Yuan namei.u...@gmail.com
This function is buggy yet tricky to debug. For now we can't pass following
script:
for i in `seq 0 2`; do
sheep/sheep -D -n -z $i -p $((7000+$i)) -c local store/$i
done
sleep 1
dog/dog cluster format -c 3
qemu-img convert linux-0.2.img sheepdog:test
On Wed, Dec 18, 2013 at 02:51:37PM +0900, Hitoshi Mitake wrote:
This update contains an important bugfix of cluster snaphsot feature. If you
are
a user of cluster snapshot, please be aware of it. It will be included in
v0.7.6-rc1.
Liu Yuan (1):
sheep: remove sha1_from_buffer()
On Wed, Dec 18, 2013 at 02:49:35PM +0900, Hitoshi Mitake wrote:
At Wed, 18 Dec 2013 13:35:21 +0800,
Liu Yuan wrote:
On Wed, Dec 18, 2013 at 01:13:15PM +0900, Hitoshi Mitake wrote:
At Wed, 18 Dec 2013 11:36:41 +0800,
Liu Yuan wrote:
On Tue, Dec 17, 2013 at 11:42:53PM +0900,
Previous protection scheme of wi-nr_thread in work.c was
unclear because wi-startup_lock was also used for protecting it
during workqueue grow/shrink. This patch let work.c protect
wi-nr_thread by the new wi-workers_lock.
Signed-off-by: Hitoshi Mitake mitake.hito...@lab.ntt.co.jp
---
lib/work.c
wi-nr_threads is protected by wi-startup_lock now, we can make the
critical section in queue_work() smaller.
Signed-off-by: Hitoshi Mitake mitake.hito...@lab.ntt.co.jp
---
lib/work.c |2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/lib/work.c b/lib/work.c
index
This patch removes confusing names in work.c:
1. before: wi-nr_workers, after: wi-nr_queued_work
2. before: struct worker_info, after: struct wq_info
Also fixes obsolete comments.
Signed-off-by: Hitoshi Mitake mitake.hito...@lab.ntt.co.jp
---
lib/work.c | 52
Current work queue mechanism of libsheepdog has some little but bother problems
related to coding style, functionality, and performance. This patchset solves
them.
v4:
- rebase on Yuan's patch
v3:
- in the 3rd patch, add a new mutex for protecting nr_threads
v2:
- preserve correct locking
On Wed, Dec 18, 2013 at 03:04:22PM +0900, Hitoshi Mitake wrote:
wi-nr_threads is protected by wi-startup_lock now, we can make the
critical section in queue_work() smaller.
Signed-off-by: Hitoshi Mitake mitake.hito...@lab.ntt.co.jp
---
lib/work.c |2 +-
1 file changed, 1 insertion(+),
At Wed, 18 Dec 2013 14:25:48 +0800,
Liu Yuan wrote:
On Wed, Dec 18, 2013 at 03:04:22PM +0900, Hitoshi Mitake wrote:
wi-nr_threads is protected by wi-startup_lock now, we can make the
critical section in queue_work() smaller.
Signed-off-by: Hitoshi Mitake mitake.hito...@lab.ntt.co.jp
On Wed, Dec 18, 2013 at 03:04:21PM +0900, Hitoshi Mitake wrote:
Previous protection scheme of wi-nr_thread in work.c was
unclear because wi-startup_lock was also used for protecting it
during workqueue grow/shrink. This patch let work.c protect
wi-nr_thread by the new wi-workers_lock.
how
At Wed, 18 Dec 2013 14:39:19 +0800,
Liu Yuan wrote:
On Wed, Dec 18, 2013 at 03:04:21PM +0900, Hitoshi Mitake wrote:
Previous protection scheme of wi-nr_thread in work.c was
unclear because wi-startup_lock was also used for protecting it
during workqueue grow/shrink. This patch let work.c
wi-nr_threads is protected by wi-startup_lock now, we can make the
critical section in queue_work() smaller.
Signed-off-by: Hitoshi Mitake mitake.hito...@lab.ntt.co.jp
---
lib/work.c |2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/lib/work.c b/lib/work.c
index
Current work queue mechanism of libsheepdog has some little but bother problems
related to coding style, functionality, and performance. This patchset solves
them.
v5:
- protect wq_need_grow() with workers_lock
v4:
- rebase on Yuan's patch
v3:
- in the 3rd patch, add a new mutex for
This patch removes confusing names in work.c:
1. before: wi-nr_workers, after: wi-nr_queued_work
2. before: struct worker_info, after: struct wq_info
Also fixes obsolete comments.
Signed-off-by: Hitoshi Mitake mitake.hito...@lab.ntt.co.jp
---
lib/work.c | 52
Previous protection scheme of wi-nr_thread in work.c was
unclear because wi-startup_lock was also used for protecting it
during workqueue grow/shrink. This patch let work.c protect
wi-nr_thread by the new wi-workers_lock.
Signed-off-by: Hitoshi Mitake mitake.hito...@lab.ntt.co.jp
---
lib/work.c
On Wed, Dec 18, 2013 at 03:43:50PM +0900, Hitoshi Mitake wrote:
At Wed, 18 Dec 2013 14:39:19 +0800,
Liu Yuan wrote:
On Wed, Dec 18, 2013 at 03:04:21PM +0900, Hitoshi Mitake wrote:
Previous protection scheme of wi-nr_thread in work.c was
unclear because wi-startup_lock was also used
At Wed, 18 Dec 2013 14:57:11 +0800,
Liu Yuan wrote:
On Wed, Dec 18, 2013 at 03:43:50PM +0900, Hitoshi Mitake wrote:
At Wed, 18 Dec 2013 14:39:19 +0800,
Liu Yuan wrote:
On Wed, Dec 18, 2013 at 03:04:21PM +0900, Hitoshi Mitake wrote:
Previous protection scheme of wi-nr_thread in
Hi sheepdog users and developers,
I released v0.7.6-rc1 of stable branch. You can download a source
archive from these URLs:
tar.gz: https://github.com/sheepdog/sheepdog/archive/v0.7.6-rc1.tar.gz
zip: https://github.com/sheepdog/sheepdog/archive/v0.7.6-rc1.zip
This update contains an important
On Wed, Dec 18, 2013 at 04:03:17PM +0900, Hitoshi Mitake wrote:
At Wed, 18 Dec 2013 14:57:11 +0800,
Liu Yuan wrote:
On Wed, Dec 18, 2013 at 03:43:50PM +0900, Hitoshi Mitake wrote:
At Wed, 18 Dec 2013 14:39:19 +0800,
Liu Yuan wrote:
On Wed, Dec 18, 2013 at 03:04:21PM +0900,
On Wed, Dec 18, 2013 at 04:03:17PM +0900, Hitoshi Mitake wrote:
At Wed, 18 Dec 2013 14:57:11 +0800,
Liu Yuan wrote:
On Wed, Dec 18, 2013 at 03:43:50PM +0900, Hitoshi Mitake wrote:
At Wed, 18 Dec 2013 14:39:19 +0800,
Liu Yuan wrote:
On Wed, Dec 18, 2013 at 03:04:21PM +0900,
58 matches
Mail list logo