Hi,
This is the expected behaviour for a distribute volume. Files that hash to a brick that is down will not be created. This is to prevent issues in case the file already exists on that brick.
To prevent this, please use distribute-replicate volumes.
Regards,
Nithya
On Thu, 8 Aug 2019 at 13:17, Nux! <nux@xxxxxxxxx> wrote:
Sorry, I meant to say distributed, not replicated!
I'm on 6.4 from CentOs7 SIG.
I was hoping the volume might still be fully usable write-wise, with
files going on the remaining bricks, but it doesn't seem to be the case.
---
Sent from the Delta quadrant using Borg technology!
On 2019-08-08 06:54, Ravishankar N wrote:
> On 07/08/19 9:53 PM, Nux! wrote:
>> Hello,
>>
>> I'm testing a replicated volume with 3 bricks. I've killed a brick,
>> but the volume is still mounted and can see the files from the bricks
>> that are still online and can do operations on them.
>> What I cannot do is create new files in the volume, e.g.:
>>
>> dd: failed to open ‘test1000’: Transport endpoint is not connected
>>
>>
>> Is there a way to make this volume continue to work while one of the
>> bricks is offline? There is still space available in the remaining
>> bricks, shouldn't it try to use it?
>
> If 2 bricks are online and the clients are connected to them, writes
> should work. Unless the brick that was down was the only good copy,
> i.e. the only one that successfully witnessed all previous writes.
> What version of gluster are you using? Check the mount log for more
> details.
>
>>
>> Regards
>>
_______________________________________________
Gluster-users mailing list
Gluster-users@xxxxxxxxxxx
https://lists.gluster.org/mailman/listinfo/gluster-users
_______________________________________________ Gluster-users mailing list Gluster-users@xxxxxxxxxxx https://lists.gluster.org/mailman/listinfo/gluster-users