[Gluster-users] Heal flapping between Possibly undergoing heal and In split brain

Thu Mar 21 11:36:00 UTC 2019

Hi Milos,

Thanks for the logs and the getfattr output.
>From the logs I can see that there are 6 entries under the
directory "/data/data-cluster/dms/final_archive" named
41be9ff5ec05c4b1c989c6053e709e59
5543982fab4b56060aa09f667a8ae617
a8b7f31775eebc8d1867e7f9de7b6eaf
c1d3f3c2d7ae90e891e671e2f20d5d4b
e5934699809a3b6dcfc5945f408b978b
e7cdc94f60d390812a5f9754885e119e
which are having gfid mismatch, so the heal is failing on this directory.

You can use the CLI option to resolve these files from gfid mismatch. You
can use any of the 3 methods available:
1. bigger-file
gluster volume heal <VOLNAME> split-brain bigger-file <FILE>

2. latest-mtime
gluster volume heal <VOLNAME> split-brain latest-mtime <FILE>

3. source-brick
gluster volume heal <VOLNAME> split-brain source-brick <HOSTNAME:BRICKNAME>
<FILE>

where <FILE> must be absolute path w.r.t. the volume, starting with '/'.
If all those entries are directories then go for either
latest-mtime/source-brick option.
After you resolve all these gfid-mismatches, run the "gluster volume heal
<volname>" command. Then check the heal info and let me know the result.

Regards,
Karthik

On Thu, Mar 21, 2019 at 4:27 PM Milos Cuculovic <cuculovic at mdpi.com> wrote:

> Sure, thank you for following up.
>
> About the commands, here is what I see:
>
> brick1:
> —————————————————————————————————————
> sudo gluster volume heal storage2 info
> Brick storage3:/data/data-cluster
> <gfid:256ca960-1601-4f0d-9b08-905c6fd52326>
> <gfid:7a63a729-c48f-4a00-9040-c3e2a0710ae6>
> /dms/final_archive - Possibly undergoing heal
>
> Status: Connected
> Number of entries: 3
>
> Brick storage4:/data/data-cluster
> <gfid:276fec9a-1c9b-4efe-9715-dcf4207e99b0>
> /dms/final_archive - Possibly undergoing heal
>
> Status: Connected
> Number of entries: 2
> —————————————————————————————————————
> sudo getfattr -d -m . -e hex /data/data-cluster/dms/final_archive
> getfattr: Removing leading '/' from absolute path names
> # file: data/data-cluster/dms/final_archive
> trusted.afr.dirty=0x000000000000000000000000
> trusted.afr.storage2-client-1=0x000000000000000000000010
> trusted.gfid=0x16c6a1e2b3fe4851972b998980097a87
> trusted.glusterfs.dht=0x000000010000000000000000ffffffff
> trusted.glusterfs.dht.mds=0x00000000
> —————————————————————————————————————
> stat /data/data-cluster/dms/final_archive
>   File: '/data/data-cluster/dms/final_archive'
>   Size: 3497984   Blocks: 8768       IO Block: 4096   directory
> Device: 807h/2055d Inode: 26427748396  Links: 72123
> Access: (0755/drwxr-xr-x)  Uid: (   33/www-data)   Gid: (   33/www-data)
> Access: 2018-10-09 04:22:40.514629044 +0200
> Modify: 2019-03-21 11:55:37.382278863 +0100
> Change: 2019-03-21 11:55:37.382278863 +0100
>  Birth: -
> —————————————————————————————————————
> —————————————————————————————————————
>
> brick2:
> —————————————————————————————————————
> sudo gluster volume heal storage2 info
> Brick storage3:/data/data-cluster
> <gfid:256ca960-1601-4f0d-9b08-905c6fd52326>
> <gfid:7a63a729-c48f-4a00-9040-c3e2a0710ae6>
> /dms/final_archive - Possibly undergoing heal
>
> Status: Connected
> Number of entries: 3
>
> Brick storage4:/data/data-cluster
> <gfid:276fec9a-1c9b-4efe-9715-dcf4207e99b0>
> /dms/final_archive - Possibly undergoing heal
>
> Status: Connected
> Number of entries: 2
> —————————————————————————————————————
> sudo getfattr -d -m . -e hex /data/data-cluster/dms/final_archive
> getfattr: Removing leading '/' from absolute path names
> # file: data/data-cluster/dms/final_archive
> trusted.afr.dirty=0x000000000000000000000000
> trusted.afr.storage2-client-0=0x000000000000000000000001
> trusted.gfid=0x16c6a1e2b3fe4851972b998980097a87
> trusted.glusterfs.dht=0x000000010000000000000000ffffffff
> trusted.glusterfs.dht.mds=0x00000000
> —————————————————————————————————————
> stat /data/data-cluster/dms/final_archive
>   File: '/data/data-cluster/dms/final_archive'
>   Size: 3497984   Blocks: 8760       IO Block: 4096   directory
> Device: 807h/2055d Inode: 13563551265  Links: 72124
> Access: (0755/drwxr-xr-x)  Uid: (   33/www-data)   Gid: (   33/www-data)
> Access: 2018-10-09 04:22:40.514629044 +0200
> Modify: 2019-03-21 11:55:46.382565124 +0100
> Change: 2019-03-21 11:55:46.382565124 +0100
>  Birth: -
> —————————————————————————————————————
>
> Hope this helps.
>
> - Kindest regards,
>
> Milos Cuculovic
> IT Manager
>
> ---
> MDPI AG
> Postfach, CH-4020 Basel, Switzerland
> Office: St. Alban-Anlage 66, 4052 Basel, Switzerland
> Tel. +41 61 683 77 35
> Fax +41 61 302 89 18
> Email: cuculovic at mdpi.com <cuculovic at mdpi.com>
> Skype: milos.cuculovic.mdpi
>
> Disclaimer: The information and files contained in this message
> are confidential and intended solely for the use of the individual or
> entity to whom they are addressed. If you have received this message in
> error, please notify me and delete this message from your system. You may
> not copy this message in its entirety or in part, or disclose its contents
> to anyone.
>
> On 21 Mar 2019, at 11:43, Karthik Subrahmanya <ksubrahm at redhat.com> wrote:
>
> Can you attach the "glustershd.log"  file which will be present under
> "/var/log/glusterfs/" from both the nodes and the "stat" & "getfattr -d -m
> . -e hex <file-path-on-brick>" output of all the entries listed in the heal
> info output from both the bricks?
>
> On Thu, Mar 21, 2019 at 3:54 PM Milos Cuculovic <cuculovic at mdpi.com>
> wrote:
>
>> Thanks Karthik!
>>
>> I was trying to find some resolution methods from [2] but unfortunately
>> none worked (I can explain what I tried if needed).
>>
>> I guess the volume you are talking about is of type replica-2 (1x2).
>>
>> That’s correct, aware of the arbiter solution but still didn’t took time
>> to implement.
>>
>> From the info results I posted, how to know in which situation I am. No
>> files are mentioned in spit brain, only directories. One brick has 3
>> entries and one two entries.
>>
>> sudo gluster volume heal storage2 info
>> [sudo] password for sshadmin:
>> Brick storage3:/data/data-cluster
>> <gfid:256ca960-1601-4f0d-9b08-905c6fd52326>
>> <gfid:7a63a729-c48f-4a00-9040-c3e2a0710ae6>
>> /dms/final_archive - Possibly undergoing heal
>>
>> Status: Connected
>> Number of entries: 3
>>
>> Brick storage4:/data/data-cluster
>> <gfid:276fec9a-1c9b-4efe-9715-dcf4207e99b0>
>> /dms/final_archive - Possibly undergoing heal
>>
>> Status: Connected
>> Number of entries: 2
>>
>> - Kindest regards,
>>
>> Milos Cuculovic
>> IT Manager
>>
>> ---
>> MDPI AG
>> Postfach, CH-4020 Basel, Switzerland
>> Office: St. Alban-Anlage 66, 4052 Basel, Switzerland
>> Tel. +41 61 683 77 35
>> Fax +41 61 302 89 18
>> Email: cuculovic at mdpi.com <cuculovic at mdpi.com>
>> Skype: milos.cuculovic.mdpi
>>
>> Disclaimer: The information and files contained in this message
>> are confidential and intended solely for the use of the individual or
>> entity to whom they are addressed. If you have received this message in
>> error, please notify me and delete this message from your system. You may
>> not copy this message in its entirety or in part, or disclose its contents
>> to anyone.
>>
>> On 21 Mar 2019, at 10:27, Karthik Subrahmanya <ksubrahm at redhat.com>
>> wrote:
>>
>> Hi,
>>
>> Note: I guess the volume you are talking about is of type replica-2
>> (1x2). Usually replica 2 volumes are prone to split-brain. If you can
>> consider converting them to arbiter or replica-3, they will handle most of
>> the cases which can lead to slit-brains. For more information see [1].
>>
>> Resolving the split-brain: [2] talks about how to interpret the heal info
>> output and different ways to resolve them using the CLI/manually/using the
>> favorite-child-policy.
>> If you are having entry split brain, and is a gfid split-brain (file/dir
>> having different gfids on the replica bricks) then you can use the CLI
>> option to resolve them. If a directory is in gfid split-brain in a
>> distributed-replicate volume and you are using the source-brick option
>> please make sure you use the brick of this subvolume, which has the same
>> gfid as that of the other distribute subvolume(s) where you have the
>> correct gfid, as the source.
>> If you are having a type mismatch then follow the steps in [3] to resolve
>> the split-brain.
>>
>> [1]
>> https://docs.gluster.org/en/v3/Administrator%20Guide/Split%20brain%20and%20ways%20to%20deal%20with%20it/
>> [2]
>> https://docs.gluster.org/en/latest/Troubleshooting/resolving-splitbrain/
>> [3]
>> https://docs.gluster.org/en/latest/Troubleshooting/resolving-splitbrain/#dir-split-brain
>>
>> HTH,
>> Karthik
>>
>> On Thu, Mar 21, 2019 at 1:45 PM Milos Cuculovic <cuculovic at mdpi.com>
>> wrote:
>>
>>> I was now able to catch the split brain log:
>>>
>>> sudo gluster volume heal storage2 info
>>> Brick storage3:/data/data-cluster
>>> <gfid:256ca960-1601-4f0d-9b08-905c6fd52326>
>>> <gfid:7a63a729-c48f-4a00-9040-c3e2a0710ae6>
>>> /dms/final_archive - Is in split-brain
>>>
>>> Status: Connected
>>> Number of entries: 3
>>>
>>> Brick storage4:/data/data-cluster
>>> <gfid:276fec9a-1c9b-4efe-9715-dcf4207e99b0>
>>> /dms/final_archive - Is in split-brain
>>>
>>> Status: Connected
>>> Number of entries: 2
>>>
>>> Milos
>>>
>>> On 21 Mar 2019, at 09:07, Milos Cuculovic <cuculovic at mdpi.com> wrote:
>>>
>>> Since 24h, after upgrading from 4.0 to 4.1.7 one of the servers, the
>>> heal shows this:
>>>
>>> sudo gluster volume heal storage2 info
>>> Brick storage3:/data/data-cluster
>>> <gfid:256ca960-1601-4f0d-9b08-905c6fd52326>
>>> <gfid:7a63a729-c48f-4a00-9040-c3e2a0710ae6>
>>> /dms/final_archive - Possibly undergoing heal
>>>
>>> Status: Connected
>>> Number of entries: 3
>>>
>>> Brick storage4:/data/data-cluster
>>> <gfid:276fec9a-1c9b-4efe-9715-dcf4207e99b0>
>>> /dms/final_archive - Possibly undergoing heal
>>>
>>> Status: Connected
>>> Number of entries: 2
>>>
>>> The same files stay there. From time to time the status of the
>>> /dms/final_archive is in split brain at the following command shows:
>>>
>>> sudo gluster volume heal storage2 info split-brain
>>> Brick storage3:/data/data-cluster
>>> /dms/final_archive
>>> Status: Connected
>>> Number of entries in split-brain: 1
>>>
>>> Brick storage4:/data/data-cluster
>>> /dms/final_archive
>>> Status: Connected
>>> Number of entries in split-brain: 1
>>>
>>> How to know the file who is in split brain? The files in
>>> /dms/final_archive are not very important, fine to remove (ideally resolve
>>> the split brain) for the ones that differ.
>>>
>>> I can only see the directory and GFID. Any idea on how to resolve this
>>> situation as I would like to continue with the upgrade on the 2nd server,
>>> and for this the heal needs to be done with 0 entries in sudo gluster
>>> volume heal storage2 info
>>>
>>> Thank you in advance, Milos.
>>>
>>>
>>> _______________________________________________
>>> Gluster-users mailing list
>>> Gluster-users at gluster.org
>>> https://lists.gluster.org/mailman/listinfo/gluster-users
>>
>>
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.gluster.org/pipermail/gluster-users/attachments/20190321/56c9c821/attachment.html>