why rbd journaling first write journal to journal pool and then write data

why rbd journaling first write journal to journal pool and then write data

[Jaze Lee]

Hi all:
   I found that rbd journaling first write journal and then write data.
   Why we not first write data and then write journal? Is there some
considerration
   for which is first ?

   Thanks a lot.

-- 
谦谦君子

Re: why rbd journaling first write journal to journal pool and then write data

[Mykola Golub]

Jaze,

On Mon, Aug 22, 2016 at 04:46:04PM +0800, Jaze Lee wrote:

>    I found that rbd journaling first write journal and then write data.
>    Why we not first write data and then write journal? Is there some
> considerration
>    for which is first ?

It looks like you already asked this question some time ago. Have you
missed Jason's response?

http://www.spinics.net/lists/ceph-devel/msg30167.html

Or are you unsatisfied with his answer?

-- 
Mykola Golub

Re: why rbd journaling first write journal to journal pool and then write data

[Ricardo Dias]

Hi Jaze,

If you write data first, and you have a failure after writing data and before writing to the journal, then you cannot recover that write operation and it will be lost.

You can see the journal as a primary replica of data.

Ricardo Dias

Sent from my mobile phone

> On 22 Aug 2016, at 09:46, Jaze Lee <jazeltq@gmail.com> wrote:
> 
> Hi all:
>   I found that rbd journaling first write journal and then write data.
>   Why we not first write data and then write journal? Is there some
> considerration
>   for which is first ?
> 
>   Thanks a lot.
> 
> -- 
> 谦谦君子
> --
> To unsubscribe from this list: send the line "unsubscribe ceph-devel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> 

Re: why rbd journaling first write journal to journal pool and then write data

[Ning Yao]

Hi, Ricardo

I got what you mean.

First of all, if we enable the rbd journaling, currently  write
latency will be doubled since it should wait for finish of the write
operation.

So If the issue is what you mentioned above. I think we have some
other ways to solve it. For example, we can snapshot the primary image
and run a whole image scan to find those objects which is different
from each other. And then delete the snapshot on the primary and
continue to replying journal again.  In this way, we can also maintain
the image consistency?
Regards
Ning Yao


2016-08-22 17:31 GMT+08:00 Ricardo Dias <ricardo.dias@suse.com>:
> Hi Jaze,
>
> If you write data first, and you have a failure after writing data and before writing to the journal, then you cannot recover that write operation and it will be lost.
>
> You can see the journal as a primary replica of data.
>
> Ricardo Dias
>
> Sent from my mobile phone
>
>> On 22 Aug 2016, at 09:46, Jaze Lee <jazeltq@gmail.com> wrote:
>>
>> Hi all:
>>   I found that rbd journaling first write journal and then write data.
>>   Why we not first write data and then write journal? Is there some
>> considerration
>>   for which is first ?
>>
>>   Thanks a lot.
>>
>> --
>> 谦谦君子
>> --
>> To unsubscribe from this list: send the line "unsubscribe ceph-devel" in
>> the body of a message to majordomo@vger.kernel.org
>> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>>
>
> --
> To unsubscribe from this list: send the line "unsubscribe ceph-devel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html