[Gluster-users] Healing issue

Ravishankar N ravishankar at redhat.com
Mon Aug 17 08:07:03 UTC 2015



On 08/17/2015 12:42 PM, Miloš Kozák wrote:
> I assumed that those files are different because of different file size.
>

I then reckon the writes are not successful on the 2nd node. The brick 
log should give some clue. You can also check the fuse mount log for errors.

> Ok, so my MD5s are:
>
> [root at nodef01i fs]# md5sum *
> 6bc6e89a9cb46a4098db300ff6d95b46 f1607f25aa52f4fb6f98f20ef0f3f9d7
> 0989d12cc875387827ebbbdc67503f2b 3706a2cb0bb27ba5787b3c12388f4ebb
>
> [root at nodef01i fs]# md5sum *
> 0989d12cc875387827ebbbdc67503f2b 3706a2cb0bb27ba5787b3c12388f4ebb
> md5sum: 6bc6e89a9cb46a4098db300ff6d95b46: Adresář nebo soubor 
> neexistuje (File or folder does not exist)
>
>
The md5sum was run on the back-end bricks right? - In which case the 
ENOENT error seems to indicate a corrupt file system. (`dmesg` might 
show something).

>
>
> Dne 17.8.2015 v 08:40 Ravishankar N napsal(a):
>>
>>
>> On 08/17/2015 11:58 AM, Miloš Kozák wrote:
>>> Mounted by fuse.
>>> I would not say written to only one node because both nodes contain 
>>> something, but I am quite suspicious about files on the node2.
>> Not sure I follow. Is there a mismatch of the file's contents on both 
>> nodes? You could do an md5sum check.
>



More information about the Gluster-users mailing list