Re: [Gluster-devel] In what kind of circumstance, the changlog trusted.afr.xxx of a file will become 0xFFFFFFFF

[Krutika Dhananjay <kdhananj@xxxxxxxxxx>]

---

From: "Krutika Dhananjay" <kdhananj@xxxxxxxxxx>
To: "Jaden Liang" <jaden1q84@xxxxxxxxx>
Cc: gluster-users@xxxxxxxxxxx
Sent: Friday, November 21, 2014 8:46:42 AM
Subject: Re: [Gluster-devel] In what kind of circumstance, the changlog trusted.afr.xxx of a file will become 0xFFFFFFFF

Changing cc'd group to gluster-users.

Hi,

Could you provide the client and brick logs (bricks 8 and 9 when counting from 0)?

-Krutika

---

From: "Jaden Liang" <jaden1q84@xxxxxxxxx>
To: gluster-devel@xxxxxxxxxxx
Sent: Friday, November 21, 2014 8:16:42 AM
Subject: [Gluster-devel] In what kind of circumstance, the changlog trusted.afr.xxx of a file will become 0xFFFFFFFF

Hi all,

I have a glusterfs-3.4.5 build with 6 x 2 Distributed-Replicate volume for KVM storage. And found one of the file is not in a consistant state. I checked the extent attributes of every replica file on brick as below:

# file: sf/data/vs/local/d2c2bf42-0206-43db-824b-d2d3872ea42d/98d0e5d0-44c6-4284-ae09-aa0a232056c7/images/cluster/b6da49e44cec/test.vm/vm-disk-1.qcow2

trusted.afr.vs_vol_rep2-client-8=0xffffffff0000000000000000

trusted.afr.vs_vol_rep2-client-9=0x000000000000000000000000

trusted.gfid=0xca90124fa56d42bba7a2840719ed74d6

trusted.glusterfs.lockinfo=0x0000000100000075000000093c504f534958282f73662f646174612f76732f6c6f63616c2f34313831616363352d633532332d343033662d383664352d3033346566656337316434652f39386430653564302d343463362d343238342d616530392d616130613233323035366337293a686f73742d34633732623965366461376500313733313236313600

# file: sf/data/vs/local/4181acc5-c523-403f-86d5-034efec71d4e/98d0e5d0-44c6-4284-ae09-aa0a232056c7/images/cluster/b6da49e44cec/test.vm/vm-disk-1.qcow2

trusted.afr.vs_vol_rep2-client-8=0x000000000000000000000000

trusted.afr.vs_vol_rep2-client-9=0x000000000000000000000000

trusted.gfid=0xca90124fa56d42bba7a2840719ed74d6

trusted.glusterfs.lockinfo=0x0000000100000075000000093c504f534958282f73662f646174612f76732f6c6f63616c2f34313831616363352d633532332d343033662d383664352d3033346566656337316434652f39386430653564302d343463362d343238342d616530392d616130613233323035366337293a686f73742d34633732623965366461376500313733313236313600

There is a trusted.afr.vs_vol_rep2-client-8=0xffffffff0000000000000000. BTW, one of the servers did disconnect for less than 1 minute then came back 1 miute and keep this loop for one 12 hours(Just testing the glusterfs behavior in bad network circumstance). As I know, the changelog of aft-xlator is modified by increase or decrease with 1. I think it is unlikely the operations increase to 0xFFFFFFFF. It might be some buggy code decrease changelog to -1 which is 0xFFFFFFFF.

And yes, to answer your question, this does seem like a bug, where a post-op of -1 on the changelog whose value was 0 before could have led to a value of 0xffffffff.

-Krutika

Just sending this mail here, may be there is someone met this before or knowing what is going on.

Thanks.

_______________________________________________
Gluster-devel mailing list
Gluster-devel@xxxxxxxxxxx
http://supercolony.gluster.org/mailman/listinfo/gluster-devel

_______________________________________________
Gluster-users mailing list
Gluster-users@xxxxxxxxxxx
http://supercolony.gluster.org/mailman/listinfo/gluster-users