[Gluster-devel] Using geo-replication as backup solution using gluster volume snapshot!

[Aravinda]

regards
Aravinda

On 03/09/2016 06:22 PM, Shyam wrote:
> On 03/09/2016 12:45 AM, Aravinda wrote:
>>
>> regards
>> Aravinda
>>
>> On 03/08/2016 11:34 PM, Atin Mukherjee wrote:
>>>
>>> On 03/07/2016 05:13 PM, Kotresh Hiremath Ravishankar wrote:
>>>> Added gluster-users.
>>>>
>>>> Thanks and Regards,
>>>> Kotresh H R
>>>>
>>>> ----- Original Message -----
>>>>> From: "Kotresh Hiremath Ravishankar" <khiremat at redhat.com>
>>>>> To: "Gluster Devel" <gluster-devel at gluster.org>
>>>>> Sent: Monday, March 7, 2016 3:03:08 PM
>>>>> Subject: [Gluster-devel] Using geo-replication as backup solution
>>>>> using gluster volume snapshot!
>>>>>
>>>>> Hi All,
>>>>>
>>>>> Here is the idea, we can use geo-replication as backup solution
>>>>> using gluster
>>>>> volume
>>>>> snapshots on slave side. One of the drawbacks of geo-replication is
>>>>> that it's
>>>>> a
>>>>> continuous asynchronous replication and would not help in getting
>>>>> the last
>>>>> week's or
>>>>> yesterday's data. So if we use gluster snapshots at the slave end,
>>>>> we can use
>>>>> the
>>>>> snapshots to get the last week's or yesterday's data making it a
>>>>> candidate
>>>>> for a
>>>>> backup solution. The limitation is that the snapshots at the slave
>>>>> end can't
>>>>> be
>>>>> restored as it will break the running geo-replication. It could be
>>>>> mounted
>>>>> and
>>>>> we have access to data when the snapshots are taken. It's just a
>>>>> naive idea.
>>>>> Any suggestions and use cases are worth discussing:)
>>> When you mention that gluster snapshot can be taken at the slave end 
>>> how
>>> does it guarantee that the data is available till yesterday or next 
>>> week
>>> considering the asynchronous nature of the replication strategy? (I've
>>> very limited knowledge on geo replication, so I may sound stupid!)
>> Check my response in the same thread. Geo-rep now has a scheduler
>> script, which can be used to run Geo-replication whenever required.
>> It does the following to make sure everything is synced to Slave.
>>
>> 1. Stop Geo-replication if Started
>> 2. Start Geo-replication
>> 3. Set Checkpoint
>> 4. Check the Status and see Checkpoint is Complete.(LOOP)
>> 5. If checkpoint complete, Stop Geo-replication
>
> Will this stop at the moment checkpoint is complete, or is there a 
> chance that geo-rep would continue with the next change log, but be 
> interrupted by this script.
Checkpoint here is just a stop notification for Scheduler script. 
Geo-rep may sync some files which are created/modified after the 
checkpoint time. But it makes sure that everything created/modified 
before checkpoint are in sync.
>
> IOW, we are polling for the checkpoint, which when we detect has 
> happened, may not mean geo-rep has not processed (or started 
> processing) the next changelog, would this understanding be right?
It may process next changelogs.
>
> I state/ask this, as we may want to geo-rep upto a checkpoint, which 
> is a point of snapshot on the master, and not geo-rep beyond this point.
Geo-rep can't provide point of snapshot since data changes are not 
recorded in changelogs. File may have modified after it recorded data in 
Changelog.
>
> I guess I need to understand checkpoints better, if my understanding 
> is incorrect.
Checkpoint is kind of indicator that shows all the files 
created/modified before the checkpoint time are synced.
>
>>>>>
>>>>> Thanks and Regards,
>>>>> Kotresh H R
>>>>>
>>>>> _______________________________________________
>>>>> Gluster-devel mailing list
>>>>> Gluster-devel at gluster.org
>>>>> http://www.gluster.org/mailman/listinfo/gluster-devel
>>>>>
>>>> _______________________________________________
>>>> Gluster-devel mailing list
>>>> Gluster-devel at gluster.org
>>>> http://www.gluster.org/mailman/listinfo/gluster-devel
>>>>
>>> _______________________________________________
>>> Gluster-devel mailing list
>>> Gluster-devel at gluster.org
>>> http://www.gluster.org/mailman/listinfo/gluster-devel
>>
>> _______________________________________________
>> Gluster-devel mailing list
>> Gluster-devel at gluster.org
>> http://www.gluster.org/mailman/listinfo/gluster-devel
> _______________________________________________
> Gluster-devel mailing list
> Gluster-devel at gluster.org
> http://www.gluster.org/mailman/listinfo/gluster-devel