SV: SV: [Linux-cluster] Linux Clustering Newbe

[Kristoffer Lippert]

Hi,
 
A fence device is a device that can "build a fence" around a node, and thus keep it from corrupting a shared filesystem.
Most commonly i think a powerswitch is used. It simply cuts the power to the decfunct server.
It looks and works like this:
http://www.wti.com/guides/rpb115ug.htm
(but there are ofcourse loads of brands avaliabel - No, I don't work for WTI ;-)
 
Alternatively there is the option of "cutting" the network from the server (that would be the network between server and disks), using a switch. I have not tried that method of fencing, so someone else might be able to fill in. 

Depending on the topology of your cluster (are your disks connected thrugh a dedicated fiber net or for instance iscsi.) the need to "cut" the defective server off from the disks will be different.
 
So in short, A fence device can be many things. :-)
 
Hope it make sence
/Kristoffer
 

________________________________

Fra: linux-cluster-bounces at redhat.com [mailto:linux-cluster-bounces at redhat.com] På vegne af Dan.Askew at jmsmucker.com
Sendt: 20. juli 2007 14:26
Til: linux clustering
Emne: Re: SV: [Linux-cluster] Linux Clustering Newbe



Could you elaborate on the fence device.  What would suggest using ? 






"Kristoffer Lippert" <kristoffer.lippert at jppol.dk> 
Sent by: linux-cluster-bounces at redhat.com 

07/20/2007 08:13 AM 
Please respond to
linux clustering <linux-cluster at redhat.com>


To
"linux clustering" <linux-cluster at redhat.com> 
cc
Subject
SV: [Linux-cluster] Linux Clustering Newbe

	




Hi, 
  
You need gfs for the changes to appear on both servers. With GFS, when one server changes a file, the other server is made aware of the changes. Also GFS takes care of file locking. Also you need a fencedevice, so the cluster can shutdown a "defective" server, and make sure it dosn't corrupt the GFS. 
  
For your current setup: 
When you have both servers running, you could mount the ext3 fs on both servers, but only the server that writes a file, will be aware of it. The other server will be aware of the new file when you remount the fs. 
  
Hope this helps a bit. 
  
Kind Regards 
Kristoffer 
  
  


________________________________

Fra: linux-cluster-bounces at redhat.com [mailto:linux-cluster-bounces at redhat.com] På vegne af Dan.Askew at jmsmucker.com
Sendt: 20. juli 2007 14:01
Til: linux clustering
Emne: [Linux-cluster] Linux Clustering Newbe


Greetings all, 

I am an old veteran of HP-UX Service Guard. I am trying to get a NFS linux Cluster working and need some advise. 

I have read the NFS Cookbook from Redhat and have a the following working 

2 Node Linux Cluster  (RHEL AS 4.0 update 5) 

one test disk LVM formated ext3 (have not decided onGFS or not) 

Use Vitual IPAddress to access the disks via NFS 

When SYSTEMA  runs the service and the client machine access the disk and makes changes. Then I fail over to SYSTEMB these changes made by the client are not present. 

I am runing CLVMD deamon 
The LVM disks are mounted on both systems 
I have made the following changes to LVM.CONF 

(I have tried locking_type = 2 and locking_type = 3) both have he same results. (as above) 

Sorry for my ignorance but can anyone tell me what I am doing wrong...would GFS solve the syncing problem? 




--
Linux-cluster mailing list
Linux-cluster at redhat.com
https://www.redhat.com/mailman/listinfo/linux-cluster 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listman.redhat.com/archives/linux-cluster/attachments/20070720/5e788b80/attachment.htm>