[Gluster-devel] How to accelerate the self-heal process?

[hao sun]

I have a replication volume as bellows:



[root at gluster001 disk]# gluster volume info rep



Volume Name: rep

Type: Replicate

Volume ID: 0f7eb8a0-92d7-4911-b766-f23ed870cf3d

Status: Started

Number of Bricks: 1 x 2 = 2

Transport-type: tcp

Bricks:

Brick1: gluster001:/disk/sdc

Brick2: gluster002:/disk/sdb



The sdc and sdb are both raid5 device in 3 SAS disks.



And I have 3 clients
(gluster-client001,gluster-client002,gluster-client003) mount on the same
replication volume.To test the self-heal function, I clean
gluster001:/disk/sdc (rm ¨Crf /disk/sdc/*), and run ¡°gluster volume heal rep
full¡±. I use the iftop to check the process and I get the log :

 [image: ÄÚǶͼƬ 2]The clients are 172.18.4.17, 172.18.4.19 and 172.18.4.20.
There are no jobs running on the clients. But from the log, the client 4.17
and 4.19 are reading something from the gluster002 server. And the
self-heal process¡¯s performance seems to be affected, only 253Mb/s.



My Questions, thanks:

@1: why/what are the clients 4.17 and 4.19 reading from gluster002?
@2: how to accelerate the self-heal?
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://supercolony.gluster.org/pipermail/gluster-devel/attachments/20130815/a33548d4/attachment-0001.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: ???? 2013-08-13 ??12.45.51.png
Type: image/png
Size: 85798 bytes
Desc: not available
URL: <http://supercolony.gluster.org/pipermail/gluster-devel/attachments/20130815/a33548d4/attachment-0001.png>