On Sun, 17 Mar 2019 10:31:20 +0900,
Miles Fidelman wrote:
>
> On 3/16/19 8:01 PM, MORITA Kazutaka wrote:
>
> > On Sun, 17 Mar 2019 08:20:25 +0900,
> > Miles Fidelman wrote:
> >> So who is using Sheepdog in production?
> > Our subsidiaries are using it.
&g
On Sun, 17 Mar 2019 09:15:52 +0900,
Raymond Burkholder wrote:
>
> On 2019-03-15 11:29 p.m., MORITA Kazutaka wrote:
> >> Any suggestions on alternatives? I'd say ceph is a primary one, but
> >> sheepdog seemed to be 'lighter weight'.
> >
> > I'm a bit inte
On Sun, 17 Mar 2019 08:20:25 +0900,
Miles Fidelman wrote:
>
> So who is using Sheepdog in production?
Our subsidiaries are using it.
Thanks,
Kazutaka
--
sheepdog mailing list
sheepdog@lists.wpkg.org
https://lists.wpkg.org/mailman/listinfo/sheepdog
Hello,
Sheepdog is not maintained now. There is no plan to add new features,
and the repository is kept only for users who are using sheepdog in
production currently.
> Any suggestions on alternatives? I'd say ceph is a primary one, but sheepdog
> seemed to be 'lighter weight'.
I'm a bit
We forked the sheepdog project. Let's leave the maintance of the original tree
to Hitoshi.
Signed-off-by: MORITA Kazutaka morita.kazut...@gmail.com
---
MAINTAINERS | 8 ++--
1 file changed, 2 insertions(+), 6 deletions(-)
diff --git a/MAINTAINERS b/MAINTAINERS
index 2688829..49dd651 100644
At Mon, 16 Mar 2015 21:13:29 +0800,
Liu Yuan wrote:
How about make 'dog vdi clone --no-share' as the default clone operation? And
we can add dog vdi clone --share to keep old behavior as optional. By this
manner, --no-share will save us from this kind of subtle problem. And your
team
I'm not working for NTT lab now.
Also, let's add Hitoshi to the master branch maintainers. Unfortunately, I
have no time for sheepdog development currently. He is the maintainer of the
stable branch and will do a good job instead of me.
Signed-off-by: MORITA Kazutaka morita.kazut...@gmail.com
At Fri, 4 Jul 2014 16:45:23 +0200,
Valerio Pachera wrote:
Has this patch been merged in the end?
Is sheepdog using ssse3 and avx to speed up checksum calculation?
Yes.
https://github.com/sheepdog/sheepdog/commit/04396f5e668f9a3bd13431e7a239eb305cb27f00
Thanks,
Kazutaka
--
sheepdog
At Tue, 3 Jun 2014 00:08:54 +0900,
Hitoshi Mitake wrote:
The object reclaim doesn't support hypervolume yet. But hypervolume cannot be
used as a virtual disk (both of qemu and tgt don't support it) currently. And
the removal of old vdi deletion is acceptable for hypervolume because it
At Mon, 17 Mar 2014 21:43:25 +0800,
Liu Yuan wrote:
With new algorithm,
$ dog vdi create image
$ dog vdi snapshot image -s snap1
$ dog vdi clone -s snap1 image clone
$ dog vdi delete clone -- this operation will surprise you that it
won't
At Wed, 26 Feb 2014 14:43:33 +0800,
Robin Dong wrote:
In our test environment, we upload 6TB data and kill one node, then the sheep
daemons on each server will try to rename files from 'data' to '.stale', but
the rename operations cost almost half an hour. And in this half an hour, the
whole
At Tue, 18 Mar 2014 00:02:18 +0800,
Liu Yuan wrote:
Keep old algorithm is a bottom line to me. Either sheep option or static
#ifdef
looks fine to me. We can later refine it to be dynamically pluggable.
New algorithm looks to me more a partial solution than a generic one, compared
with
At Mon, 24 Feb 2014 11:08:28 +0800,
Liu Yuan wrote:
On Sun, Feb 23, 2014 at 07:08:27PM +0900, Hitoshi Mitake wrote:
On Sun, Feb 23, 2014 at 2:28 PM, Hitoshi Mitake
mitake.hito...@gmail.com wrote:
The object reclaim doesn't support hypervolume yet. But hypervolume
cannot be
used as
At Thu, 6 Feb 2014 17:08:49 +0800,
Liu Yuan wrote:
Signed-off-by: Liu Yuan namei.u...@gmail.com
---
sheep/recovery.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
Applied, thanks.
Kazutaka
--
sheepdog mailing list
sheepdog@lists.wpkg.org
At Thu, 06 Feb 2014 14:37:25 +0900,
Hitoshi Mitake wrote:
At Mon, 3 Feb 2014 10:05:49 +0900,
Hitoshi Mitake wrote:
Signed-off-by: Hitoshi Mitake mitake.hito...@lab.ntt.co.jp
---
sheep/vdi.c |2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
ping?
Applied, thanks.
At Thu, 06 Feb 2014 14:36:38 +0900,
Hitoshi Mitake wrote:
At Thu, 6 Feb 2014 01:34:18 +0800,
Liu Yuan wrote:
If we need it future, we can bring them back on demand from git repo. No
need
to keep it in the source file.
Signed-off-by: Liu Yuan namei.u...@gmail.com
---
At Thu, 06 Feb 2014 14:34:35 +0900,
Hitoshi Mitake wrote:
At Wed, 5 Feb 2014 21:17:28 +0800,
Liu Yuan wrote:
- pass correct array size
- return FULL_VDI instead of wrong NO_BASE_VDI
Signed-off-by: Liu Yuan namei.u...@gmail.com
---
sheep/vdi.c | 12 ++--
1 file
At Thu, 06 Feb 2014 14:22:40 +0900,
Hitoshi Mitake wrote:
At Wed, 5 Feb 2014 19:36:58 +0900,
Ryusuke Konishi wrote:
Get rid of the following untracked files from the output of git
status command:
$ git status
# On branch install
# Untracked files:
# (use git add
At Thu, 6 Feb 2014 17:09:42 +0900,
Yoshinori Matsuo wrote:
Sheep output the end of auto-recovery log message (all nodes are recovered)
at DEBUG level.
When sheep nodes are working in the INFO level or upper,
the operator could not get this message
even if operators want to know the
At Wed, 5 Feb 2014 15:41:25 +0800,
Liu Yuan wrote:
+
+ /*
+ * If running_threads_nr == 0, it means run_next_rw() is called in the
+ * finish_object_list() and no recovery threads so it is safe to run
+ * into next recovery.
+ *
+ * If running_threads_nr = 1,
At Thu, 6 Feb 2014 17:18:57 +0800,
Liu Yuan wrote:
Rationale for multi-threaded recovery:
1. If one node is added, we find that all the VMs on other nodes will get
noticeably affected until 50% data is transferred to the new node.
2. For node failure, we might not have problems of
At Tue, 4 Feb 2014 15:14:44 +0800,
Liu Yuan wrote:
@@ -666,12 +671,19 @@ static void free_recovery_info(struct recovery_info
*rinfo)
/* Return true if next recovery work is queued. */
static inline bool run_next_rw(void)
{
- struct recovery_info *nrinfo =
At Tue, 4 Feb 2014 14:40:14 +0100,
Marcin Miros$(D)H(Baw wrote:
Signed-off-by: Marcin Miros$(D)H(Baw mar...@mejor.pl
---
sheep/sheep_priv.h | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
Applied, thanks.
Kazutaka
--
sheepdog mailing list
sheepdog@lists.wpkg.org
At Tue, 4 Feb 2014 11:53:43 +0800,
Liu Yuan wrote:
This is a defensive patch, but solve a problem we found on the test cluster.
The
problem is tricky:
init_vdi_state() errorr out as failed to read inode header xxx, 0.
After a detailed log analyzing, we found that the cause of this
At Tue, 28 Jan 2014 18:01:42 +0800,
Liu Yuan wrote:
@@ -861,6 +885,7 @@ static void recover_object_main(struct work *work)
finish_recovery(rinfo);
out:
+ rinfo-running_threads_nr--;
free_recovery_obj_work(row);
}
This can cause a segfault because rinfo may be freed in
Signed-off-by: MORITA Kazutaka morita.kazut...@lab.ntt.co.jp
---
sheep/recovery.c |4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/sheep/recovery.c b/sheep/recovery.c
index c15297c..50b6c58 100644
--- a/sheep/recovery.c
+++ b/sheep/recovery.c
@@ -613,8 +613,8 @@ main_fn
We cannot access dw after we call delete_vdis_done() because it may be
already freed.
Signed-off-by: MORITA Kazutaka morita.kazut...@lab.ntt.co.jp
---
sheep/vdi.c |8
1 file changed, 4 insertions(+), 4 deletions(-)
diff --git a/sheep/vdi.c b/sheep/vdi.c
index 3414b05..080bc82
After all the parities are set to out[], p will increase beyond
ctx-dp. This adds a check for it.
This also adds assert() to make sure that we don't overrun the buffer.
Signed-off-by: MORITA Kazutaka morita.kazut...@lab.ntt.co.jp
---
lib/fec.c |3 ++-
1 file changed, 2 insertions(+), 1
This silences valgrind errors.
Signed-off-by: MORITA Kazutaka morita.kazut...@lab.ntt.co.jp
---
sheep/object_cache.c |1 +
1 file changed, 1 insertion(+)
diff --git a/sheep/object_cache.c b/sheep/object_cache.c
index 09d3950..d51d459 100644
--- a/sheep/object_cache.c
+++ b/sheep
There is no guarantee that req-data is a string. Actually, the
current code can cause a buffer overrun when, e.g.,
SD_OP_FORCE_RECOVER is requested.
Signed-off-by: MORITA Kazutaka morita.kazut...@lab.ntt.co.jp
---
include/util.h |1 +
lib/util.c | 16
sheep
req-op can be NULL when the requested opcode is invalid. It's safe
to add NULL checks before accessing req-op in ops.c.
Signed-off-by: MORITA Kazutaka morita.kazut...@lab.ntt.co.jp
---
sheep/ops.c | 19 +++
1 file changed, 11 insertions(+), 8 deletions(-)
diff --git a/sheep
At Fri, 31 Jan 2014 13:31:29 +0900,
Matsuo Yoshinori wrote:
Hitoshi I discussed again.
We have an another future idea that adds context id in the log message to
trace easily.
i.e.
dog command logs:
Jan 31 13:19:34 sheep-01-00 sheepdog admin operation[1998]: dog cluster
format,
At Fri, 31 Jan 2014 14:11:06 +0900,
Yoshinori Matsuo wrote:
Administrator couldn't track operation on each node without this function.
This function logs only administrative operation at rx_main and tx_main.
Operations that is not related to administrative operation are not
logged in this
At Thu, 30 Jan 2014 12:47:49 +0800,
Liu Yuan wrote:
@@ -254,6 +267,11 @@ int default_init(void)
return for_each_object_in_wd(init_objlist_and_vdi_bitmap, true, NULL);
}
+static inline bool stale_path(const char *path)
+{
+ return !!strstr(path, stale);
+}
+
The function
At Thu, 30 Jan 2014 12:47:46 +0800,
Liu Yuan wrote:
This patch set fixes the problem
'init_vdi_state() errorr out as failed to read inode header xxx, 0.'
at sheep refuses to restart due to misplacement object but the fix would be
more
broader to fix all the misplacement problems if
At Fri, 31 Jan 2014 10:18:21 +0900,
Hitoshi Mitake wrote:
At Thu, 30 Jan 2014 14:05:45 +0800,
Liu Yuan wrote:
Signed-off-by: Liu Yuan namei.u...@gmail.com
---
tests/functional/032 | 1 +
1 file changed, 1 insertion(+)
Looks good to me.
Reviewed-by: Hitoshi Mitake
At Wed, 29 Jan 2014 16:46:40 +0900,
Hitoshi Mitake wrote:
Especially new VID of current VDI is useful information. So this patch
lets dog print both of the IDs.
Signed-off-by: Hitoshi Mitake mitake.hito...@lab.ntt.co.jp
---
dog/vdi.c |9 +
1 file changed, 5 insertions(+), 4
Signed-off-by: MORITA Kazutaka morita.kazut...@lab.ntt.co.jp
---
tests/functional/common.rc |2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/tests/functional/common.rc b/tests/functional/common.rc
index 713136a..5388e48 100644
--- a/tests/functional/common.rc
+++ b/tests
| 31 +++---
3 files changed, 81 insertions(+), 37 deletions(-)
Reviewed-by: MORITA Kazutaka morita.kazut...@lab.ntt.co.jp
Thanks,
Kazutaka
--
sheepdog mailing list
sheepdog@lists.wpkg.org
http://lists.wpkg.org/mailman/listinfo/sheepdog
At Wed, 29 Jan 2014 11:55:29 +0900,
Hitoshi Mitake wrote:
At Tue, 28 Jan 2014 16:53:56 +0900,
MORITA Kazutaka wrote:
At Tue, 28 Jan 2014 11:34:22 +0800,
Liu Yuan wrote:
On Mon, Jan 27, 2014 at 07:54:35PM +0900, Hitoshi Mitake wrote:
This patch adds a new parameter dst
At Tue, 28 Jan 2014 19:30:39 +0900,
Hitoshi Mitake wrote:
At Tue, 28 Jan 2014 17:04:14 +0900,
MORITA Kazutaka wrote:
At Tue, 28 Jan 2014 15:26:52 +0900,
Hitoshi Mitake wrote:
Only Kazutaka knows the reason. I guess it was planned to be paralle. I
think
deletion
At Wed, 29 Jan 2014 17:52:40 +0900,
Hitoshi Mitake wrote:
At Wed, 29 Jan 2014 16:26:14 +0800,
Liu Yuan wrote:
v3:
- fix a gcc warning when nfs isn't enabled
Now the problems I pointed out seem to be resolved. Thanks for your
rework. This patchset looks good to me as a start point
At Tue, 28 Jan 2014 15:26:52 +0900,
Hitoshi Mitake wrote:
Only Kazutaka knows the reason. I guess it was planned to be paralle. I
think
deletion code is really dirty and tricky and hard to maintain, need
refactor and
redesign.
Of course I agree so I'm writing this patch.
At Tue, 28 Jan 2014 15:02:02 +0900,
Hitoshi Mitake wrote:
At Tue, 28 Jan 2014 11:13:34 +0800,
Liu Yuan wrote:
Forgot to remove -lcurl in my last patch, remove it.
Signed-off-by: Liu Yuan namei.u...@gmail.com
---
sheepfs/Makefile.am | 2 +-
1 file changed, 1 insertion(+), 1
At Tue, 28 Jan 2014 11:34:22 +0800,
Liu Yuan wrote:
On Mon, Jan 27, 2014 at 07:54:35PM +0900, Hitoshi Mitake wrote:
This patch adds a new parameter dst= to the -l option for
specifying logging destination. Possible parameters are below:
- default: a dedicated file under sheep's directory
At Mon, 27 Jan 2014 12:02:02 +0800,
Liu Yuan wrote:
Signed-off-by: Liu Yuan namei.u...@gmail.com
---
configure.ac| 8 +---
sheepfs/Makefile.am | 11 ++-
sheepfs/core.c | 2 ++
sheepfs/sheepfs.h | 5 +
4 files changed, 22 insertions(+), 4 deletions(-)
At Thu, 23 Jan 2014 13:57:50 +0800,
Robin Dong wrote:
From: Robin Dong san...@taobao.com
Some old softwares can only use files on filesystem as input, so we should add
a new type of interface for sheepfs to access object-storage as a
posix-filesystem.
It is called http interface, since it
At Thu, 23 Jan 2014 15:04:19 +0800,
Liu Yuan wrote:
also update 081.out for a wrong update by commit 71c86fa
Signed-off-by: Liu Yuan namei.u...@gmail.com
---
tests/functional/029.out | 16 ++--
tests/functional/030.out | 44 -
At Thu, 23 Jan 2014 15:06:35 +0800,
Liu Yuan wrote:
Remove it fix a segfault on ubuntu-13.04. Though I don't know why.
Signed-off-by: Liu Yuan namei.u...@gmail.com
---
tests/functional/common.rc | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
Applied, thanks.
Kazutaka
--
At Fri, 24 Jan 2014 12:53:15 +0900,
Matsuo Yoshinori wrote:
Wrapping dog command can also solve dog logging,
but dog command may be called directory especially in the small cluster.
And it's simpler that having a logging function in dog command
than writing log on every wrapper scripts,
At Wed, 22 Jan 2014 14:04:35 +0800,
Liu Yuan wrote:
diff --git a/tests/functional/common.rc b/tests/functional/common.rc
index 3785bae..b304580 100644
--- a/tests/functional/common.rc
+++ b/tests/functional/common.rc
@@ -486,7 +486,7 @@ _cluster_format()
_random()
{
- openssl
At Wed, 22 Jan 2014 17:41:18 +0800,
Liu Yuan wrote:
On Wed, Jan 22, 2014 at 06:18:52PM +0900, Hitoshi Mitake wrote:
At Tue, 21 Jan 2014 20:38:07 +0900,
Matsuo Yoshinori wrote:
I think the both logging function are needed because of the reasons below.
Reason:
1. The dog and
At Thu, 23 Jan 2014 14:08:23 +0800,
Liu Yuan wrote:
On Thu, Jan 23, 2014 at 02:45:48PM +0900, MORITA Kazutaka wrote:
At Wed, 22 Jan 2014 14:04:35 +0800,
Liu Yuan wrote:
diff --git a/tests/functional/common.rc b/tests/functional/common.rc
index 3785bae..b304580 100644
At Fri, 10 Jan 2014 15:05:21 +0900,
Yoshinori Matsuo wrote:
Administrator couldn't track operation on each node without this function.
This function logs only administrative operation at rx_main and tx_main.
Operations that is not related to administrative operation are not
logged in this
Hi all,
I'm pleased to announce the availability of the 0.8.0 release.
The software is available from GitHub:
https://github.com/sheepdog/sheepdog/archive/v0.8.0.tar.gz
The changes from the previous release are available at:
https://github.com/sheepdog/sheepdog/blob/v0.8.0/CHANGELOG.md
At Wed, 20 Nov 2013 14:04:43 +0800,
Robin Dong wrote:
+
+/* dump the information of B-tree */
+static void dump_btree(read_node_fn reader, struct sd_inode *inode)
+{
+#ifdef DEBUG
+ sd_info(btree BEGIN);
+ traverse_btree(reader, inode, dump_cb, NULL);
+ sd_info(btree END);
At Tue, 14 Jan 2014 17:46:26 +0900,
Hitoshi Mitake wrote:
@@ -620,6 +607,15 @@ static void logger(char *log_dir, char *outfile)
unblock_sighup();
+ if (getppid() == 1) {
+ /* My parent (sheep process) is dead. */
+
At Fri, 17 Jan 2014 16:32:51 +0800,
Liu Yuan wrote:
On Fri, Jan 17, 2014 at 12:35:31PM +0900, MORITA Kazutaka wrote:
From: MORITA Kazutaka morita.kazut...@lab.ntt.co.jp
We can use our work queue framework for handling asynchrouns requests.
There is no need to create short living events
At Thu, 16 Jan 2014 16:32:42 +0800,
Liu Yuan wrote:
Since current code works well, do we really need to introduce prioritized
event
loop which need some overhead of sorting events?
For now with this patch set, the only benefit is that it looks more
'beautiful'
that we have complete
At Thu, 16 Jan 2014 15:25:40 +0800,
Liu Yuan wrote:
This will reduce the unecessary recovery procedures if multiple node events
happens.
Signed-off-by: Liu Yuan namei.u...@gmail.com
---
sheep/recovery.c | 30 ++
1 file changed, 22 insertions(+), 8
At Thu, 16 Jan 2014 22:06:18 +0800,
Liu Yuan wrote:
On Thu, Jan 16, 2014 at 10:58:35PM +0900, MORITA Kazutaka wrote:
At Thu, 16 Jan 2014 16:32:42 +0800,
Liu Yuan wrote:
Since current code works well, do we really need to introduce prioritized
event
loop which need some
From: MORITA Kazutaka morita.kazut...@lab.ntt.co.jp
We can implement asynchronous local requests on our work queue framework.
MORITA Kazutaka (2):
sheep: use work queue for asynchronous local request
Revert sheep: allow {register,unregister}_event to be called in
worker thread
include
From: MORITA Kazutaka morita.kazut...@lab.ntt.co.jp
We can use our work queue framework for handling asynchrouns requests.
There is no need to create short living events.
Signed-off-by: MORITA Kazutaka morita.kazut...@lab.ntt.co.jp
---
sheep/request.c| 73
From: MORITA Kazutaka morita.kazut...@lab.ntt.co.jp
Now there is no user who calls register_event in worker threads.
This reverts commit 6601e90cf2c5168dfc15919caabcaef50eb75483.
Signed-off-by: MORITA Kazutaka morita.kazut...@lab.ntt.co.jp
---
include/event.h | 15 ---
lib/event.c
At Wed, 15 Jan 2014 18:12:42 +0900,
Hitoshi Mitake wrote:
At Tue, 14 Jan 2014 11:05:10 +0900,
Hitoshi Mitake wrote:
At Sun, 12 Jan 2014 19:59:51 +0900,
Hitoshi Mitake wrote:
On my ubuntu box (13.04), the openssl command called by _urandom()
dies with segmentation fault (yes,
At Wed, 15 Jan 2014 16:54:42 +0900,
Hitoshi Mitake wrote:
At Wed, 15 Jan 2014 15:16:00 +0800,
Liu Yuan wrote:
This actuall fixes a bug that 045 fails with -ec specified. Instead of add a
memset() to init_erasure_buffer() to fix the bug, I prefer to make xvalloc()
a zeroed memory
At Wed, 15 Jan 2014 13:40:07 +0800,
Liu Yuan wrote:
Allocate the buffer size based on the request length so that small request
won't
take too much memory.
Signed-off-by: Liu Yuan namei.u...@gmail.com
---
sheep/http/kv.c | 12 +++-
1 file changed, 7 insertions(+), 5
At Wed, 15 Jan 2014 14:37:15 +0800,
Liu Yuan wrote:
This will reduce the unecessary recovery procedures if multiple node events
happens.
Signed-off-by: Liu Yuan namei.u...@gmail.com
---
sheep/recovery.c | 28
1 file changed, 20 insertions(+), 8 deletions(-)
At Mon, 16 Dec 2013 09:05:04 +0100,
Valerio Pachera wrote:
Hi all, is there any news about sheepdog 0.8 release date?
Sorry for my too late reply. I've been very busy for other works.
I wanted to add a feature to reclaim unused objects before the
release, but I'll release sheepdog 0.8 asap
Hi all,
I'd like to announce the release candidate for sheepdog 0.8.0.
https://github.com/sheepdog/sheepdog/archive/v0.8.0-rc0.tar.gz
Any kind of testing would be highly appreciated. Sheepdog 0.8.0 is
planed to be released one week later:
From: MORITA Kazutaka morita.kazut...@lab.ntt.co.jp
We cannot specify a number as a log level.
Signed-off-by: MORITA Kazutaka morita.kazut...@lab.ntt.co.jp
---
sheep/sheep.c | 36 ++--
1 file changed, 18 insertions(+), 18 deletions(-)
diff --git a/sheep
At Mon, 6 Jan 2014 17:16:22 +0800,
Robin Dong wrote:
Hi All,
At present, the implemention of swift interface for creating object in
sheepdog is:
1. lock container
2. check whether the onode with same object name is exists.
3. unlock container
4. upload object
5. create onode
this
At Wed, 8 Jan 2014 09:34:00 +0800,
Robin Dong wrote:
At Mon, 6 Jan 2014 17:16:22 +0800,
Robin Dong wrote:
Hi All,
At present, the implemention of swift interface for creating object in
sheepdog is:
1. lock container
2. check whether the onode with same object name is
At Thu, 12 Dec 2013 13:45:33 +0800,
Liu Yuan wrote:
We make sure we write the exact number of copies to honor the promise of the
redundancy for strict mode. This means that after writing of targeted data,
they are redundant as promised and can withstand the random node failures.
For
At Thu, 12 Dec 2013 14:37:23 +0800,
Liu Yuan wrote:
Ah, yes. I think SD_FLAG_CMD_WRITE is kind of misleading (couple of times
I
forgot to assigne it while write something and spent time debugging it)
I think so too. hdr.data_length != 0 may have same meaning to the
flag. It
At Wed, 27 Nov 2013 13:28:36 +0800,
Robin Dong wrote:
Maybe we can add -lock for cluster framework in sheepdog and just
implement zookeeper -lock but leave corosync -lock empty.
Will this be better?
I prefer the approach.
Thanks,
Kazutaka
--
sheepdog mailing list
sheepdog@lists.wpkg.org
At Tue, 26 Nov 2013 16:38:01 +0800,
Robin Dong wrote:
Getting a distributed lock is an expensive operation and it can causes
a severe performance problem if we do it for each object creation.
Can we find another way? Sheepdog is not designed to allow concurrent
write access.
It
At Mon, 25 Nov 2013 15:03:46 +0800,
Robin Dong wrote:
The present implementation of http/swift is not perfect, it can't create
too much containers or objects. So we want to store all objects in one
hyper volume vdi and use new structure 'obj-inode' to identify its offset
and length in this
At Mon, 25 Nov 2013 17:02:06 +0800,
Liu Yuan wrote:
On Mon, Nov 25, 2013 at 05:43:19PM +0900, MORITA Kazutaka wrote:
At Mon, 25 Nov 2013 15:03:46 +0800,
Robin Dong wrote:
The present implementation of http/swift is not perfect, it can't create
too much containers or objects. So we
At Sat, 23 Nov 2013 02:30:32 +0800,
Liu Yuan wrote:
We have stable branch concept and I guess a simply copy of
get_maintainer.pl from other project won't work out. Might need much work
to tailor the script to our needs. I'd like to leave it to a future patch.
Okay, applied thanks.
At Wed, 20 Nov 2013 19:36:10 +0800,
Liu Yuan wrote:
- use sd_inode helper to update inode correctly
- pass system copy_policy when creating bucket
Signed-off-by: Liu Yuan namei.u...@gmail.com
---
-set 'direct' as false for sd_inode_write_vid()
sheep/http/kv.c | 20
: Files and directories with wildcard patterns
+
+Sheepdog Overall
+--
+M: MORITA Kazutaka morita.kazut...@lab.ntt.co.jp
+M: Liu Yuan namei.u...@gmail.com
+B: sheeepdog/*
+L: sheepdog@lists.wpkg.org
+
+Sheepdog Stable Branches
+--
+M
At Fri, 15 Nov 2013 14:19:34 +0800,
Liu Yuan wrote:
There is currently no client using the rsp-obj.copies. Since this assignment
is in the hottest code path, it is beneficial to remove it.
Signed-off-by: Liu Yuan namei.u...@gmail.com
---
sheep/ops.c |5 -
1 file changed, 5
At Sat, 16 Nov 2013 22:57:29 +0800,
Robin Dong wrote:
Hi all,
The size of vdi can only reach 4TB beacause the inode-data_vdi_id[] can
only
support 1 million objects. But 4TB is too small for storage application
such as NAS and cloud-disk therefore we need to change the array of
At Tue, 12 Nov 2013 16:14:20 +0800,
Liu Yuan wrote:
Use fprintf directly as we need
What's the benefit? sd logger is designed to be able to use fprintf()
for foreground programs. In addition, we cannot completely decouple
logger from dog anyway because dog can call sd_emerg() in panic().
At Thu, 14 Nov 2013 13:45:47 +0800,
Liu Yuan wrote:
This patch enables 'cluster format -c x:y' option for users to set a
cluster-wide default scheme.
Signed-off-by: Liu Yuan namei.u...@gmail.com
---
dog/cluster.c| 25 +
dog/common.c |
At Thu, 14 Nov 2013 17:03:53 +0800,
Liu Yuan wrote:
This will remove the unnecessary setting conent length for 'get' operation
Signed-off-by: Liu Yuan namei.u...@gmail.com
---
sheep/http/http.c | 10 ++
1 file changed, 6 insertions(+), 4 deletions(-)
diff --git
At Thu, 14 Nov 2013 17:03:54 +0800,
Liu Yuan wrote:
Signed-off-by: Liu Yuan namei.u...@gmail.com
---
sheep/http/http.c | 22 ++
1 file changed, 6 insertions(+), 16 deletions(-)
diff --git a/sheep/http/http.c b/sheep/http/http.c
index 644cdc5..e9afedc 100644
---
At Thu, 14 Nov 2013 17:03:52 +0800,
Liu Yuan wrote:
Current code will segfault if we don't specify DOCUMENT_URI in nginx conf
file.
And we should use REQUEST_URI because REQUEST_URI includes the Query String
(?q=testing...) where DOCUMENT_URI does not. ? is needed for container list
At Fri, 1 Nov 2013 16:31:00 +0800,
Liu Yuan wrote:
diff --git a/dog/common.c b/dog/common.c
index 4480b99..c243386 100644
--- a/dog/common.c
+++ b/dog/common.c
@@ -435,3 +435,41 @@ void dump_loglevels(bool err)
sd_info(%s\t(%d), loglevel_table[i], i);
}
}
+
At Fri, 1 Nov 2013 16:31:01 +0800,
Liu Yuan wrote:
Signed-off-by: Liu Yuan namei.u...@gmail.com
---
tests/functional/077 | 27 +++
tests/functional/077.out | 11 +++
tests/functional/group |1 +
3 files changed, 39 insertions(+)
create mode
At Thu, 31 Oct 2013 13:49:28 +0800,
Liu Yuan wrote:
+/*
+ * Sheepdog support two kinds of redundancy, full replication and erasure
+ * coding.
+ *
+ * # create a fully replicated vdi with x copies
+ * -o redundancy=x (1 = x = SD_MAX_COPIES)
+ *
+ * # create a erasure coded vdi with x
/sheepdog.c | 127
+
include/block/block_int.h |1 +
2 files changed, 105 insertions(+), 23 deletions(-)
Reviewed-by: MORITA Kazutaka morita.kazut...@lab.ntt.co.jp
--
sheepdog mailing list
sheepdog@lists.wpkg.org
http://lists.wpkg.org/mailman
From: MORITA Kazutaka morita.kazut...@lab.ntt.co.jp
This introduces a framework of http driver for object storage feature.
We can specify several http drivers to control http requests. The
drivers are stackable and work like a WSGI pipeline of Python.
Usage
$ sheep -r [host=hostname,][port
From: MORITA Kazutaka morita.kazut...@lab.ntt.co.jp
This adds a basic support for S3 and Swift APIs.
TODO items:
- Improve compatibility with S3 and Swift.
- Improve performance of the underlying kvs feature.
- Add support for authorization.
- Add support for large objects than 4MB.
- Add
From: MORITA Kazutaka morita.kazut...@lab.ntt.co.jp
Signed-off-by: MORITA Kazutaka morita.kazut...@lab.ntt.co.jp
---
include/util.h | 3 +++
lib/util.c | 53 +
2 files changed, 56 insertions(+)
diff --git a/include/util.h b/include/util.h
From: MORITA Kazutaka morita.kazut...@lab.ntt.co.jp
Signed-off-by: MORITA Kazutaka morita.kazut...@lab.ntt.co.jp
---
sheep/http.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/sheep/http.c b/sheep/http.c
index e4c1657..9e5551e 100644
--- a/sheep/http.c
+++ b/sheep/http.c
From: MORITA Kazutaka morita.kazut...@lab.ntt.co.jp
Signed-off-by: MORITA Kazutaka morita.kazut...@lab.ntt.co.jp
---
sheep/Makefile.am | 2 +-
sheep/http/s3.c | 283 ++
2 files changed, 284 insertions(+), 1 deletion(-)
create mode 100644
From: MORITA Kazutaka morita.kazut...@lab.ntt.co.jp
These fields are necessary to implement REST APIs in the later patch.
Signed-off-by: MORITA Kazutaka morita.kazut...@lab.ntt.co.jp
---
sheep/http.c | 24 +---
1 file changed, 17 insertions(+), 7 deletions(-)
diff --git
From: MORITA Kazutaka morita.kazut...@lab.ntt.co.jp
This prepares for introducing http driver implementation.
Signed-off-by: MORITA Kazutaka morita.kazut...@lab.ntt.co.jp
---
sheep/Makefile.am | 2 +-
sheep/http.c | 76 +--
sheep/http.h
1 - 100 of 2316 matches
Mail list logo