Gluster users - Nov 2014 - heal questions

Just some basic question on the heal process, please just point me to the docs 
if they are there :)

- How is the need for a heal detected? I presume nodes can detect when they 
can't sync writes to the other nodes. This is flagged (xattr?) for healing 
when the other nodes are back up?

- How is the master file chosen? quorum? latest write?

- I have a three peer, two node/brick replica with server side quorum. The 
third peer was added to give an odd number of peers for quorom, but it
doesn't
participate in bricks. Does this make sense? it helps with deciding which node 
is master for heals?

- diff heals. I'm guessing the nodes scan the file pairs block by block, 
calculating a checksum for each block and comparing, if they don't match,
the
master node block is replicated.

- Synchronizing the compare over the network must be fun ...
  

- Is there anyway to display a file heal progress? it would make the process a 
lot less anxiety inducing even if it was very slow.

thanks,

-- 
Lindsay
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 198 bytes
Desc: This is a digitally signed message part.
URL:
<http://supercolony.gluster.org/pipermail/gluster-users/attachments/20141119/373f7430/attachment.sig>

On 11/19/2014 03:23 AM, Lindsay Mathieson wrote:
> Just some basic question on the heal process, please just point me to the
docs
> if they are there :)
> 
> - How is the need for a heal detected? I presume nodes can detect when they
> can't sync writes to the other nodes. This is flagged (xattr?) for
healing
> when the other nodes are back up?
> 
> - How is the master file chosen? quorum? latest write?
Self heal is detection/healing is handled by AFR using extended attributes which
keep track of the type of heal (data/meta-data/entry) and the direction of heal
(i.e. what the sources and sinks are).
More details can be found here
:https://github.com/gluster/glusterfs/blob/master/doc/features/afr-v1.md

> 
> - I have a three peer, two node/brick replica with server side quorum. The 
> third peer was added to give an odd number of peers for quorom, but it
doesn't
> participate in bricks. Does this make sense? it helps with deciding which
node
> is master for heals?
Server quorum is useful for detecting split-brains due to graph change, etc but
I guess it can still lead to data split-brains. Client quorum can be used to
avoid data split-brains at the cost of
availability. More information here:

https://access.redhat.com/documentation/en-US/Red_Hat_Storage/2.1/html/Administration_Guide/ch10s10.html
https://access.redhat.com/documentation/en-US/Red_Hat_Storage/2.0/html/Administration_Guide/sect-User_Guide-Managing_Volumes-Quorum.html
http://www.gluster.org/community/documentation/index.php/Features/Server-quorum


> 
> - diff heals. I'm guessing the nodes scan the file pairs block by
block,
> calculating a checksum for each block and comparing, if they don't
match, the
> master node block is replicated.
That is correct. It scans in 128 kb chunks (configurable via the
`cluster.self-heal-window-size` volume set option).
> 
> - Synchronizing the compare over the network must be fun ...
>   
> 
> - Is there anyway to display a file heal progress? it would make the
process a
> lot less anxiety inducing even if it was very slow.
As of now, `gluster volume heal <volname> info` gives only the list of
files that need healing. But yes, it would be nice to have some sort of a CLI to
monitor the % progress of files that are
currently under data self-heal.


HTH,
Ravi
> 
> thanks,
> 
> 
> 
> _______________________________________________
> Gluster-users mailing list
> Gluster-users at gluster.org
> http://supercolony.gluster.org/mailman/listinfo/gluster-users
>