[Gluster-users] md5sum-s of file on different nodes are different.

Ilya Galanin iamdpoke at gmail.com
Fri Aug 28 06:19:17 UTC 2009


Hello Everyone.
I am using gluster 2.0.6 on my ftp-cluster on RHEL5. It contains 3 nodes.

This is package list from one node. Other nodes have the same packages 
installed.
===============================================
[ftp1:~]:999# rpm -qa | grep gluster
glusterfs-common-2.0.6-1.el5.x86_64
glusterfs-server-2.0.6-1.el5.x86_64
glusterfs-client-2.0.6-1.el5.x86_64
[ftp1:~]:1000# rpm -qa | grep fuse
fuse-libs-2.7.4glfs11-1.x86_64
kernel-module-fuse-2.6.18-128.1.1.el5-2.7.4glfs11-1.x86_64
fuse-2.7.4glfs11-1.x86_64
fuse-devel-2.7.4glfs11-1.x86_64
===============================================

My configuration files:
Server config:
===============================================
volume posix
  type storage/posix
  option directory /home/automat
end-volume

volume locks
  type features/locks
  subvolumes posix
end-volume

volume brick
  type performance/io-threads
  option thread-count 8
  subvolumes locks
end-volume

volume server
  type protocol/server
  option transport-type tcp
  option auth.addr.brick.allow *
  subvolumes brick
end-volume
===============================================

Client Config:
===============================================
# file: /etc/glusterfs/glusterfs-client.vol
#
volume remote1
  type protocol/client
  option transport-type tcp
  option remote-host 10.7.69.145
  option remote-subvolume brick
end-volume

volume remote2
  type protocol/client
  option transport-type tcp
  option remote-host 10.7.69.146
  option remote-subvolume brick
end-volume

volume remote3
  type protocol/client
  option transport-type tcp
  option remote-host 10.7.69.147
  option remote-subvolume brick
end-volume

volume replicate
  type cluster/replicate
  subvolumes remote1 remote2 remote3
end-volume

volume distribute
  type cluster/distribute
  subvolumes replicate
end-volume

volume writebehind
  type performance/write-behind
  option block-size 128KB
  option cache-size 100MB
  subvolumes distribute
end-volume

volume cache
  type performance/io-cache
  option cache-size 1024MB
  subvolumes writebehind
end-volume
===============================================

Also I have a perl script that check all files in some folders of 
ftp-server every 10 minutes and
caclulate their control sum, then write it in xml format to one new file 
on ftp (index.xml)

PS: I run this script on node1 using cron.

Sometimes after deploying some new files on ftp server (node one).
I see that md5sum of file(index.xml) is different from one node to others.
Usually node one give right md5sum. Others give wrong ... or one of 
others gives wrong ....

May be someone see that different cluster nodes give different files?
Any other suggestions and advices are welcome too.

PS. Excuse me for my English, I am not a native speaker.

Thanks in advance.
Ilya.



More information about the Gluster-users mailing list