Pranith Kumar Karampuri
2016-Sep-22 13:50 UTC
[Gluster-users] gluster 3.7 healing errors (no data available, buf->ia_gfid is null)
On Thu, Sep 22, 2016 at 12:51 PM, Ravishankar N <ravishankar at redhat.com> wrote:> On 09/22/2016 12:38 PM, Pasi K?rkk?inen wrote: > >> On Thu, Sep 22, 2016 at 09:58:25AM +0530, Ravishankar N wrote: >> >>> On 09/21/2016 10:54 PM, Pasi K?rkk?inen wrote: >>> >>>> Let's see. >>>> >>>> # getfattr -m . -d -e hex /bricks/vol1/brick1/foo >>>> getfattr: Removing leading '/' from absolute path names >>>> # file: bricks/vol1/brick1/foo >>>> security.selinux=0x756e636f6e66696e65645f753a6f626a6563745f7 >>>> 23a756e6c6162656c65645f743a733000 >>>> >>>> So hmm.. no trusted.gfid it seems.. is that perhaps because this node >>>> was down when the file was created? >>>> >>> No, even if that were the case, the gfid should have been set while >>> healing the file to this node. >>> Can you try doing a setfattr -n trusted.gfid -v >>> 0xc1ca778ed2af4828b981171c0c5bd45e on the file. and launch heal >>> again? >>> What about the .glusterfs hardlink- does that exist? >>> >>> It seems there's no hardlink.. nothing in /bricks/vol1/brick1/.glusterfs/c1/ca/ >> directory. >> >> Now I manually set the trusted.gfid value on the file, and launched heal >> again, >> and now gluster was able to heal it OK! Healing is now fully complete, >> and no out-of-sync files anymore. >> >> Any idea what caused the missing trusted.gfid ? >> >Do you want to raise a bug for this? We would love to if you don't have the time to make sure we address this.> A create FOP is a multi step process on the bricks. > 1.creating the file on the actual path > 2. Setting the gluster xattrs including gfid xattr > 3. Creating the link file inside .glusterfs > > I'm guessing your brick went down after step 1 for the files in question. > Check the brick logs to check for such messages. If the brick was still up, > check if there are logs for failures related to performing 2 and 3. > > By the way, if everything healed successfully, check that the .glusterfs > hardlink is now present. > > -Ravi > > > >> >> >> Thanks a lot! >> >> -- Pasi >> >> > _______________________________________________ > Gluster-users mailing list > Gluster-users at gluster.org > http://www.gluster.org/mailman/listinfo/gluster-users >-- Pranith -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://www.gluster.org/pipermail/gluster-users/attachments/20160922/a726558d/attachment.html>
Pasi Kärkkäinen
2016-Sep-22 15:03 UTC
[Gluster-users] gluster 3.7 healing errors (no data available, buf->ia_gfid is null)
On Thu, Sep 22, 2016 at 07:20:26PM +0530, Pranith Kumar Karampuri wrote:> On Thu, Sep 22, 2016 at 12:51 PM, Ravishankar N > <[1]ravishankar at redhat.com> wrote: > > On 09/22/2016 12:38 PM, Pasi K??rkk??inen wrote: > > On Thu, Sep 22, 2016 at 09:58:25AM +0530, Ravishankar N wrote: > > On 09/21/2016 10:54 PM, Pasi K??rkk??inen wrote: > > Let's see. > > # getfattr -m . -d -e hex /bricks/vol1/brick1/foo > getfattr: Removing leading '/' from absolute path names > # file: bricks/vol1/brick1/foo > security.selinux=0x756e636f6e66696e65645f753a6f626a6563745f723a756e6c6162656c65645f743a733000 > > So hmm.. no trusted.gfid it seems.. is that perhaps because this > node was down when the file was created? > > No, even if that were the case, the gfid should have been set while > healing the file to this node. > Can you try doing a setfattr -n trusted.gfid -v > 0xc1ca778ed2af4828b981171c0c5bd45e on the file. and launch heal > again? > What about the .glusterfs hardlink- does that exist? > > It seems there's no hardlink.. nothing in > /bricks/vol1/brick1/.glusterfs/c1/ca/ directory. > > Now I manually set the trusted.gfid value on the file, and launched > heal again, > and now gluster was able to heal it OK! Healing is now fully complete, > and no out-of-sync files anymore. > > Any idea what caused the missing trusted.gfid ? > > Do you want to raise a bug for this? We would love to if you don't have > the time to make sure we address this. >Sure. Should I file the bug on redhat bugzilla? -- Pasi