[Gluster-users] File deletion gets corrupted on cluster-resize.

Raghavendra Bhat raghavendrabhat at gluster.com
Fri Sep 9 02:35:29 PDT 2011


On Thu, Sep 8, 2011 at 11:14 PM, Kazuyoshi Tlacaelel <kazu.dev at gmail.com>wrote:

> Thanks for your feedback.
>
> You can find the logs here. http://cl.ly/273p0T0E1r2X2C243M2s
>
> gluster> volume info all
>
> Volume Name: gluster_replication
> Type: Distributed-Replicate
> Status: Started
> Number of Bricks: 2 x 2 = 4
> Transport-type: tcp
> Bricks:
> Brick1: 110.1.1.256:/ebs/gluster_2_volume_1
> Brick2: 110.2.2.256:/ebs/gluster_1_volume_1
> Brick3: 110.3.3.256:/ebs/gluster_2_volume_2
> Brick4: 110.4.4.256:/ebs/gluster_1_volume_2
>
>
> The error happens when you perform the resize (adding: gluster_2_volume_2
> & gluster_1_volume_2)
> and run the following command at the same time.
>
> cp -r rails /gluster/external_copy & cp -r /gluster/original
>> /gluster/internal_copy & rm -rf /gluster/delete
>>
>> In the above command, what is rails? In the description you have not
mentioned anything about rails. Can please provide information about it?


>
>
>
> On Sep 8, 2011, at 2:32 AM, Raghavendra Bhat wrote:
>
> Hi Kazuyoshi Tlacaelel,
>
> Can you please provide the glusterfs client, glusterfs server and glusterd
> logs? They will be present in /usr/local/var/log/glusterfs for source
> installation (if you hace installed with some prefix, then logs will be
> present in <prefix_path>/var/log/glusterfs). If you have installed from rpm,
> then logs will be present in /var/log/glusterfs.
>
> Also can you provide the output of gluster volume info <volname>?
>
> Thanks
>
> Regards,
> Raghavendra Bhat
>
> On Thu, Sep 8, 2011 at 2:49 AM, Kazuyoshi Tlacaelel <kazu.dev at gmail.com>wrote:
>
>> Two master-servers, one client.
>>
>> *Client mount point is*: */gluster*
>> *replica is*: *2*
>>
>> A *testing-directory* (git repo) with data exists, and is used for
>> testing the following.
>>
>> *STEP 1*
>>
>>    - First, we copy the* testing-directory* to /gluster/original
>>    - Then, we copy the *testing-directory* to /gluster/delete
>>
>> *STEP 2*
>>
>>    - Copy /gluster/original to /gluster/internal_copy
>>    - Copy *testing-directory* to /gluster/external_copy
>>    - Delete /gluster/delete
>>
>> cp -r rails /gluster/external_copy & cp -r /gluster/original
>> /gluster/internal_copy & rm -rf /gluster/delete
>>
>> While executing the command above, only the delete part returns an error:
>>
>> *rm: FATAL: directory `/gluster/delete/activerecord/test/migrations'
>> changed dev/ino*
>>
>> *Note*: that this corruption only happens while the a new-brick is added
>> to the system.
>> and that that is executed at the same time as *STEP 2*.
>> Which purpose is to replicate a live production environment.
>>
>>
>> *RESULT*
>>
>> I've confirmed that the files within the following directories are
>> consistent:
>>
>>
>>    - /gluster/original
>>    - /gluster/external_copy
>>    - /gluster/internal_copy
>>
>> But delete doesn't get deleted, its actually just "*partially-deleted*".
>> then, when everything is done, and a delete is performed manually the
>> directory gets deleted just fine.
>>
>> *SYSTEM*
>>
>>
>> *fuse init (API version 7.13)* Haven't done any gluster-recommended
>> fuse-compilation
>> *glusterfs 3.2.3 built on Aug 23 2011 18:54:07 *Error is replicable in
>> this version.
>> *glusterfs 3.1.2 built on Jan 18 2011 11:19:54 *Error is replicable in
>> this version.
>> *Ubuntu 10.4 64bit *All tests were performed in this machines.
>>
>>
>> _______________________________________________
>> Gluster-users mailing list
>> Gluster-users at gluster.org
>> http://gluster.org/cgi-bin/mailman/listinfo/gluster-users
>>
>>
>
>
Thanks

Regards,
Raghavendra Bhat
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://gluster.org/pipermail/gluster-users/attachments/20110909/c75c69ea/attachment.htm>


More information about the Gluster-users mailing list