On 12/05/2011 08:07 PM, anthony garnier wrote:
thank you for  your answer.
A last question for you, on some directory I got a little difference when I display extended attribute :

getfattr -d -m "trusted*" -e hex *

Server1 # file: yvask300
trusted.gfid=0x433c796513864672871226072825336f
trusted.glusterfs.dht=0x00000001000000007fffffffffffffff

Server2 # file: yvask300
trusted.gfid=0x433c796513864672871226072825336f
trusted.glusterfs.dht=0x00000001000000007fffffffffffffff


Server3 # file: yvask300
trusted.afr.poolsave-client-2=0x000000000000000002000000
trusted.afr.poolsave-client-3=0x000000000000000002000000
trusted.gfid=0x433c796513864672871226072825336f
trusted.glusterfs.dht=0x0000000100000000000000007ffffffe

Server4 # # file: yvask300
trusted.afr.poolsave-client-2=0x000000000000000002000000
trusted.afr.poolsave-client-3=0x000000000000000002000000
trusted.gfid=0x433c796513864672871226072825336f
trusted.glusterfs.dht=0x0000000100000000000000007ffffffe

why the attribute "trusted.afr.VOLNAME-client-X" is not displayed on Server 1 & 2 ?


Thx for your help.

------------------------------------------------------------------------
Date: Fri, 2 Dec 2011 08:59:48 +0530
From: [email protected]
To: [email protected]
CC: [email protected]; [email protected]
Subject: Re: [Gluster-users] Detecting split brain

On 12/01/2011 11:33 PM, anthony garnier wrote:

    So if I understand well, a value different than 0sAAAAAAAAAAAAAAAA
    for the attribute trusted.afr.poolsave-client-0 indicate split brain.

    Thx

    > Date: Thu, 1 Dec 2011 11:23:29 -0500
    > From: [email protected] <mailto:[email protected]>
    > To: [email protected] <mailto:[email protected]>
    > CC: [email protected] <mailto:[email protected]>
    > Subject: Re: [Gluster-users] Detecting split brain
    >
    > On Thu, 1 Dec 2011 16:00:09 +0000
    > anthony garnier <[email protected]>
    <mailto:[email protected]> wrote:
    >
    > >
    > > Hi,
    > >
    > > I got a lot of files with attributes : 0sAAAAAAAAAAAAAAAAA
    > >
    > > Serv 1 :
    > > # file: tbo_rmr_globale_log_11-04-07_15h43m48s.log
    > > trusted.afr.poolsave-client-0=0sAAAAAAAAAAAAAAAA
    > > trusted.afr.poolsave-client-1=0sAAAAAAAAAAAAAAAA
    > > trusted.gfid=0sfm7CRROuQ4+wuQfmHjFCdg==
    > >
    > > Serv 2 :
    > > # file: tbo_rmr_globale_log_11-04-07_15h43m48s.log
    > > trusted.afr.poolsave-client-0=0sAAAAAAAAAAAAAAAA
    > > trusted.afr.poolsave-client-1=0sAAAAAAAAAAAAAAAA
    > > trusted.gfid=0sfm7CRROuQ4+wuQfmHjFCdg==
    > >
    > > Does it mean that those files need Self-Healing ? I use GlusterFS
    > > 3.2.3
    >
    > This is actually normal. For some reason that would probably make me
    > throw up if I knew it, getfattr misreports
    0x000000000000000000000000
    > as 0sAAAAAAAAAAAAAAAA (which would be 0x404040404040404040404040) if
    > you don't give it the "-e hex" flag. The value is actually three
    > four-byte integers, and if they're zero it means there are no
    pending
    > operations. Any *other* value is likely to indicate split brain.


    _______________________________________________
    Gluster-users mailing list
    [email protected]  <mailto:[email protected]>
    http://gluster.org/cgi-bin/mailman/listinfo/gluster-users

Small Correction:
Different value than 0sAAA.. means there are pending operations.
If you dump the values with -e hex option and there should be 24 digits, first 8 represent pending operations on the data of the files (write/truncate). next 8 represent pending operations on metadata (permissions, ownership etc) next 8 represent pending operations on entry (creation/deletion/rename of a file inside that directory). Now if both data/metadata/entry digits are non-zero on a file then that will be split-brain.
Example:
Data split-brain:
 trusted.afr.poolsave-client-0=0s000000010000000000000000
 trusted.afr.poolsave-client-1=0s000000200000000000000000

Thanks
Pranith

The xattrs are set on need basis so no xattr means all zeros.

Thanks
Pranith
_______________________________________________
Gluster-users mailing list
[email protected]
http://gluster.org/cgi-bin/mailman/listinfo/gluster-users

Reply via email to