Re: [sheepdog] [PATCH v4] sheepdog: selectable object size support

(2015/02/12 16:42), Liu Yuan wrote: On Thu, Feb 12, 2015 at 04:28:01PM +0900, Hitoshi Mitake wrote: At Thu, 12 Feb 2015 15:00:49 +0800, Liu Yuan wrote: On Thu, Feb 12, 2015 at 03:19:21PM +0900, Hitoshi Mitake wrote: At Tue, 10 Feb 2015 18:35:58 +0800, Liu Yuan wrote: On Tue, Feb 10, 2015 at

Re: [sheepdog] [PATCH v4] sheepdog: selectable object size support

On Thu, Feb 12, 2015 at 05:01:05PM +0900, Teruaki Ishizaki wrote: > (2015/02/12 16:42), Liu Yuan wrote: > >On Thu, Feb 12, 2015 at 04:28:01PM +0900, Hitoshi Mitake wrote: > >>At Thu, 12 Feb 2015 15:00:49 +0800, > >>Liu Yuan wrote: > >>> > >>>On Thu, Feb 12, 2015 at 03:19:21PM +0900, Hitoshi Mitake

Re: [sheepdog] [PATCH v4] sheepdog: selectable object size support

At Thu, 12 Feb 2015 15:42:17 +0800, Liu Yuan wrote: > > On Thu, Feb 12, 2015 at 04:28:01PM +0900, Hitoshi Mitake wrote: > > At Thu, 12 Feb 2015 15:00:49 +0800, > > Liu Yuan wrote: > > > > > > On Thu, Feb 12, 2015 at 03:19:21PM +0900, Hitoshi Mitake wrote: > > > > At Tue, 10 Feb 2015 18:35:58 +080

Re: [sheepdog] [PATCH 1/2] dog: add a new option for reducing identical snapshots

At Thu, 12 Feb 2015 15:49:00 +0800, Liu Yuan wrote: > > On Thu, Feb 12, 2015 at 04:40:56PM +0900, Hitoshi Mitake wrote: > > At Thu, 12 Feb 2015 15:31:15 +0800, > > Liu Yuan wrote: > > > > > > On Thu, Feb 12, 2015 at 03:59:51PM +0900, Hitoshi Mitake wrote: > > > > At Thu, 12 Feb 2015 14:38:37 +080

Re: [sheepdog] [PATCH v4] sheepdog: selectable object size support

On Thu, Feb 12, 2015 at 05:13:55PM +0900, Hitoshi Mitake wrote: > At Thu, 12 Feb 2015 15:42:17 +0800, > Liu Yuan wrote: > > > > On Thu, Feb 12, 2015 at 04:28:01PM +0900, Hitoshi Mitake wrote: > > > At Thu, 12 Feb 2015 15:00:49 +0800, > > > Liu Yuan wrote: > > > > > > > > On Thu, Feb 12, 2015 at 0

[sheepdog] [PATCH v2 2/2] tests/functional: add a test for reducing identical snapshots

This patch add a new test for the functionality of reducing identical snapshots. With -R option, new snapshot isn't created if a VDI doesn't have its own objects. Signed-off-by: Hitoshi Mitake --- tests/functional/100 | 32 tests/functional/100.out | 9 +

[sheepdog] [PATCH v2 1/2] dog: add a new option for reducing identical snapshots

Current "dog vdi snapshot" command creates a new snapshot unconditionally, even if a working VDI doesn't have its own objects. In such a case, the created snapshot is redundant because same VDI is already existing. This patch adds a new option -R to the dog command for reducing the identical snaps

Re: [sheepdog] [PATCH v3] sheepdog:show more detail of the crash source

At Thu, 12 Feb 2015 00:28:14 -0800, Wang Zhengyong wrote: > > In the current sheepdog, when sheepdog crashed, > there is too little information about the signal source. > > This patch use (*handler)(int, siginfo_t *, void *) > instead of (*handler)(int). In this way, can show more detail of the >

Re: [sheepdog] [PATCH] zookeeper: show real session timeout in zk_init()

At Wed, 11 Feb 2015 14:20:02 +0800, Liu Yuan wrote: > > From: Liu Yuan > > ZK server has its own session timeout range, so client's timeout value passed > to server might be ignored and return the value it thinks proper. > > From the documentation, the server range is > > [tickTime * 2, tickTi

Re: [sheepdog] [PATCH] zookeeper: add more detailed description on how zk_watcher report states

At Wed, 11 Feb 2015 18:42:16 +0800, Liu Yuan wrote: > > From: Liu Yuan > > Signed-off-by: Liu Yuan > --- > sheep/cluster/zookeeper.c | 25 +++-- > 1 file changed, 23 insertions(+), 2 deletions(-) Cc-ing Ruoyu. If he cannot allocate time for review, I'll do it later. Thank

Re: [sheepdog] [PATCH v2 1/2] dog: add a new option for reducing identical snapshots

On Thu, Feb 12, 2015 at 05:26:59PM +0900, Hitoshi Mitake wrote: > Current "dog vdi snapshot" command creates a new snapshot > unconditionally, even if a working VDI doesn't have its own > objects. In such a case, the created snapshot is redundant because > same VDI is already existing. > > This pa

Re: [sheepdog] obbject placement

Hi Yua, Am 11.02.2015 um 11:44 schrieb Liu Yuan: > Yes, you are right if we don't consider the virtual nodes added to consistent > hashing to mitigate the object migration > problem. But with virtual nodes in the picture, we can add one more rule to > make sure all the data and ec-chunks are > o

Re: [sheepdog] reovery and consistency questions

Hi Hitoshi, Am 10.02.2015 um 08:31 schrieb Hitoshi Mitake: > At the epoch 8 and 9, client cannot access to sheepdog because all > members of latest healthy epoch (in this case, 6) aren't gathered yet. > In such a case, you can see an output of cluster info command like > below: > > $ dog cluster i

[sheepdog] [PATCH v3] sheepdog:show more detail of the crash source

In the current sheepdog, when sheepdog crashed, there is too little information about the signal source. This patch use (*handler)(int, siginfo_t *, void *) instead of (*handler)(int). In this way, can show more detail of the crash problem, especially the pid of singal sender Cc: Hitoshi Mitake

[sheepdog] [PATCH] Make the vdi lock mechanism optional

Current sheepdog driver will send LOCK_VDI/RELEASE_VDI ops, while start/stop QEMU VM with sheepdog protocol, this mechanism will lead qemu-img/qemu-io report error when the sheepdog VDI is locking To keep backward compatiblity, this patch modify the functionality that make the vdi lock mechanism o

[sheepdog] [PATCH] sheepdog: Fix misleading error messages in sd_snapshot_create()

If do_sd_create() fails, it first reports the error returned, then reports a another one with strerror(errno). errno is meaningless at that point. Report just one error combining the valid information from both messages. Reported-by: Eric Blake Signed-off-by: Markus Armbruster --- Applies on t

[sheepdog] Several sheepdog functions mix up -1 and -errno in return values

Two common conventions for functions returning int that may fail: 1. Return non-negative value on success, -1 value on failure. 2. Return non-negative value on success, a negative errno error code on failure. Both work. But mixing them in the same function is not a good idea. Suspicious fun

Re: [sheepdog] [PATCH] Make the vdi lock mechanism optional

At Thu, 12 Feb 2015 00:53:25 -0800, Wang Zhengyong wrote: > > Current sheepdog driver will send LOCK_VDI/RELEASE_VDI ops, > while start/stop QEMU VM with sheepdog protocol, this mechanism > will lead qemu-img/qemu-io report error when the sheepdog VDI is locking > > To keep backward compatiblity,

Re: [sheepdog] [Qemu-devel] [PATCH] sheepdog: Fix misleading error messages in sd_snapshot_create()

On 02/12/2015 06:49 AM, Markus Armbruster wrote: > If do_sd_create() fails, it first reports the error returned, then > reports a another one with strerror(errno). errno is meaningless at > that point. > > Report just one error combining the valid information from both > messages. > > Reported-b

Re: [sheepdog] Redundancy policy via iSCSI

At Sat, 7 Feb 2015 10:03:18 +0800, hujianyang wrote: > > On 2015/2/6 16:41, Hitoshi Mitake wrote: > > At Wed, 04 Feb 2015 11:24:21 +0900, > > Hitoshi Mitake wrote: > >> > >> At Tue, 3 Feb 2015 17:17:42 +0800, > >> hujianyang wrote: > >>> > >>> Hi Saeki, > >>> > >>> On 2015/2/3 16:53, Saeki Masaki

[sheepdog] Build failed in Jenkins: sheepdog-build #632

See Changes: [liuyuan] dog: fix to calculate a resizable max VDI size appropriately [mitake.hitoshi] sheepdog:show more detail of the crash source [liuyuan] dog: add a new option for reducing identical snapshots [liuyuan

Re: [sheepdog] [PATCH v4] sheepdog: selectable object size support

(2015/02/12 11:55), Liu Yuan wrote: On Thu, Feb 12, 2015 at 11:33:16AM +0900, Teruaki Ishizaki wrote: (2015/02/12 11:19), Liu Yuan wrote: On Thu, Feb 12, 2015 at 10:51:25AM +0900, Teruaki Ishizaki wrote: (2015/02/10 20:12), Liu Yuan wrote: On Tue, Jan 27, 2015 at 05:35:27PM +0900, Teruaki Ish

Re: [sheepdog] [PATCH v4] sheepdog: selectable object size support

On Fri, Feb 13, 2015 at 10:33:04AM +0900, Teruaki Ishizaki wrote: > (2015/02/12 11:55), Liu Yuan wrote: > >On Thu, Feb 12, 2015 at 11:33:16AM +0900, Teruaki Ishizaki wrote: > >>(2015/02/12 11:19), Liu Yuan wrote: > >>>On Thu, Feb 12, 2015 at 10:51:25AM +0900, Teruaki Ishizaki wrote: > (2015/02/

[sheepdog] [PATCH] sheepdog: fix confused return values

From: Liu Yuan These functions mix up -1 and -errno in return values and would might cause trouble error handling in the call chain. This patch let them return -errno and add some comments. Reported-by: Markus Armbruster Signed-off-by: Liu Yuan --- block/sheepdog.c | 11 +++ 1 file c

Re: [sheepdog] [PATCH] sheepdog: fix confused return values

On Fri, Feb 13, 2015 at 11:45:42AM +0800, Liu Yuan wrote: > From: Liu Yuan > > These functions mix up -1 and -errno in return values and would might cause > trouble error handling in the call chain. > > This patch let them return -errno and add some comments. > > Reported-by: Markus Armbruster

Re: [sheepdog] [PATCH] sheepdog: Fix misleading error messages in sd_snapshot_create()

On Thu, Feb 12, 2015 at 02:49:50PM +0100, Markus Armbruster wrote: > If do_sd_create() fails, it first reports the error returned, then > reports a another one with strerror(errno). errno is meaningless at > that point. > > Report just one error combining the valid information from both > message

Re: [sheepdog] [PATCH v4] sheepdog: selectable object size support

(2015/02/13 11:01), Liu Yuan wrote: On Fri, Feb 13, 2015 at 10:33:04AM +0900, Teruaki Ishizaki wrote: (2015/02/12 11:55), Liu Yuan wrote: On Thu, Feb 12, 2015 at 11:33:16AM +0900, Teruaki Ishizaki wrote: (2015/02/12 11:19), Liu Yuan wrote: On Thu, Feb 12, 2015 at 10:51:25AM +0900, Teruaki Ish