[Gluster-devel] [Gluster-users] Self Heal and dangling symlinks

[Lalatendu Mohanty]

On 11/21/2013 07:54 PM, Alexandre Fournier wrote:
> They are both regular file on the node and the replicas and they have the same GFID.  I ran also the gluster volume heal gv0 split-brain command and the file is not in the list.  We have an entire directory though (1023 entry on a node)
>
> However, the file was already on the brick before uploading it and I noticed that that the write did not work since the last modification date does not match the upload time.
>
> Through a web service, we offer to upload files on the gluster mount.   This web service  upload the file on a temporary folder and then MOVE the file on the gluster mount.
>
> Could the move operation give strange behavior like this?
Alexandre,

No, it should not. Please let us know the answers of the questions 
Pranith and I asked, so we can understand the root cause of your problem.

>
> Alexandre Fournier
> Tools Programmer
> Ubisoft Production Services
>
>
> -----Original Message-----
> From: Pranith Kumar Karampuri [mailto:pkarampu at redhat.com]
> Sent: 21 novembre 2013 00:47
> To: Lalatendu Mohanty
> Cc: Alexandre Fournier; gluster-users at gluster.org; gluster-devel at nongnu.org
> Subject: Re: [Gluster-devel] [Gluster-users] Self Heal and dangling symlinks
>
> Alexandre,
>     Seems like there is an entry split-brain (same file/dir name but on one brick it is a file and on the other it is a directory) according to the following log:
>> [2013-11-18 18:18:43.052446] W
>> [afr-common.c:1411:afr_conflicting_iattrs]
>> 0-gv0-replicate-0: /aa/aa/aa/aa: filetype differs on subvolumes (0, 1)
> Could you get us the output of "stat <brick-dir-path>/aa/aa/aa/aa/aa" and "getfattr -d -m. -e hex <brick-dir-path>/aa/aa/aa/aa/aa" on both the bricks.
>
> Pranith
> ----- Original Message -----
>> From: "Lalatendu Mohanty" <lmohanty at redhat.com>
>> To: "Alexandre Fournier" <alexandre.fournier at ubisoft.com>,
>> gluster-users at gluster.org, gluster-devel at nongnu.org
>> Sent: Thursday, November 21, 2013 1:28:01 AM
>> Subject: Re: [Gluster-devel] [Gluster-users] Self Heal and dangling
>> symlinks
>>
>> On 11/19/2013 10:49 PM, Alexandre Fournier wrote:
>>
>>
>>
>>
>>
>> Hello,
>>
>>
>>
>> We are experiencing strange behavior when writing file on the Gluster
>> mount point. On some occasion, when writing to the Gluster Mount we
>> have an Open Stream error. We’ve looked the gluster logs and found the
>> following faulty entries :
>>
>>
>>
>> [From /var/log/glusterfs/mnt-gv0.log]
>>
>>
>>
>> [2013-11-18 18:18:43.052446] W
>> [afr-common.c:1411:afr_conflicting_iattrs]
>> 0-gv0-replicate-0: /aa/aa/aa/aa: filetype differs on subvolumes (0, 1)
>>
>> [2013-11-18 18:18:43.052468] E
>> [afr-self-heal-common.c:1409:afr_sh_common_lookup_cbk] 0-gv0-replicate-0:
>> Conflicting entries for /aa/aa/aa/aa
>>
>> [2013-11-18 18:18:43.052757] E
>> [afr-self-heal-common.c:2160:afr_self_heal_completion_cbk]
>> 0-gv0-replicate-0: background meta-data data entry missing-entry gfid
>> self-heal
>>
>> failed on /aa/aa/aa/aa/aa
>>
>> [2013-11-18 18:18:43.052780] W [fuse-bridge.c:292:fuse_entry_cbk]
>> 0-glusterfs-fuse: 439382194: LOOKUP() /aa/aa/aa/aa/aa => -1
>> (Input/output
>> error)
>>
>>
>>
>> We’ve looked at the log file etc-glusterfs-glusterd.vol.log but we
>> found nothing related to this problem. Then, we’ve looked at the log
>>  From /var/log/glusterfs/bricks/mnt-data.log and we found 70 gig of
>> logs of the same type :
>>
>>
>>
>> [2013-11-19 17:13:32.269757] W [posix-handle.c:538:posix_handle_soft]
>> 0-gv0-posix: symlink
>> ../../ab/fe/abfeb61c-501d-4417-b8fb-0accdd57146f/cf ->
>> /mnt/data/.glusterfs/ab/fe/abfeb61c-501d-4417-b8fb-0accdd57146f/cf
>> failed (No such file or directory)
>>
>> [2013-11-19 17:13:32.269978] W [posix-handle.c:538:posix_handle_soft]
>> 0-gv0-posix: symlink
>> ../../c7/8b/c78be78f-cc95-47b2-a27f-4217f1759b67/d2 ->
>> /mnt/data/.glusterfs/c7/8b/c78be78f-cc95-47b2-a27f-4217f1759b67/d2
>> failed (No such file or directory)
>>
>> [2013-11-19 17:13:32.270190] W [posix-handle.c:538:posix_handle_soft]
>> 0-gv0-posix: symlink
>> ../../5a/8f/5a8fa43c-4ccc-4d88-9122-a96bc8ffaebc/f2 ->
>> /mnt/data/.glusterfs/5a/8f/5a8fa43c-4ccc-4d88-9122-a96bc8ffaebc/f2
>> failed (No such file or directory)
>>
>>
>>
>> This looks like a bug, unless there is something wrong with the
>> set-up. I have copied gluster-devel in this thread as I think they might help.
>>
>> Just curious, is all your gluster nodes have equal time (i.e. ntp synced).
>>
>>
>>
>>
>>
>>
>> And it does not stop logging. It seems that the self heal is not
>> working properly when there are broking symlinks in the gluster. It is
>> worth saying also that this log is only produce on a single node but
>> the write fail on several node though. Also, we try to clean the
>> symlinks manually but it always come back.
>>
>>
>>
>> Is it possible to recover from broken symlinks?
>>
>>
>>
>> Configuration :
>>
>> Gluster Version : 3.3.2
>>
>> Cluster setup : 4 X 2
>>
>> OS : Ubuntu
>>
>> On Fuse
>>
>>
>>
>> Thanks,
>>
>>
>>
>> Alexandre
>>
>>
>>
>>
>> _______________________________________________
>> Gluster-users mailing list Gluster-users at gluster.org
>> http://supercolony.gluster.org/mailman/listinfo/gluster-users
>>
>>
>> _______________________________________________
>> Gluster-devel mailing list
>> Gluster-devel at nongnu.org
>> https://lists.nongnu.org/mailman/listinfo/gluster-devel
>>