There is no doc to my knowledge. I will write one, now that
you pointed it out. ;)
Client-side heals are heals that happen from the GlusterFS
client, as part of LOOKUP, sometimes even READ, [F]STAT, etc.
You can tell when a client did a heal if a corresponding
log message appears in that client's log file (usually the
ones with MSGID: 108026, although name heals aren't logged).
They are enabled by default.
To disable entry heal from clients for instance, you do
#gluster volume set <VOL> cluster.entry-self-heal off
To disable data self-heal from clients, you do #gluster
volume set <VOL> cluster.data-self-heal off
If you want to prevent the client from doing self-heal
altogether, you disable all three forms of client healing:
data-self-heal, entry-self-heal and metadata-self-heal.
Server side heals are heals that are performed by the
self-heal daemon. You will see log messages with MSGID: 108026
in glustershd.log when the self-heal-daemon performs a heal.
Server side heals happen in the following cases:
1) when you execute #gluster volume heal <VOL> full
2) when you execute #gluster volume heal <VOL>
3) when a brick that was previously down comes back up.
To disable server side heal, you need to disable the
self-heal-daemon. You can do that with #gluster volume set
<VOL> cluster.self-heal-daemon off
HTH,
Krutika
From:
"Lindsay Mathieson" <lindsay.mathieson@xxxxxxxxx>
To: "Krutika Dhananjay" <kdhananj@xxxxxxxxxx>
Cc: "gluster-users" <gluster-users@xxxxxxxxxxx>
Sent: Friday, January 22, 2016 5:51:01 PM
Subject: Re: File Corruption when
adding bricks to live replica volumes
On 22/01/2016 10:07 PM, Krutika Dhananjay wrote:
> Could you do the following:
> 1) Disable client-side healing:
Will do. Wasn't aware there were different types of healing -
is there
client and server side heals? are there any docs on this?
thanks,
--
Lindsay Mathieson
