[Linux-cluster] NFS failover problem

[Rainer Duffner]

J. Bruce Fields wrote:
> On Tue, Aug 21, 2007 at 03:45:22PM +0200, kieran JOYEUX wrote:
>   
>> For the moment the two nfs share's content are the same. I use scp to copy 
>> them. It's static so i did it just once.
>>     
>
> So, you just did something like this?:
>
> 	scp -r /shares/someshare/ otherserver:/shares/
>
>   
>> To resolve that NFS file handle error, i mounted the share with a static 
>> fsid for each client.
>>     
>
> That's not enough.
>
>   
>> Unfortunately, even with a FSID, i still have an 
>> error during the failover while copying from the NFS share to client's 
>> local disk...
>>
>> cp: reading `/usr/local/genome/1Go.txt': Input/output error
>> cp: reading `/usr/local/genome/1Go.tyt': Input/output error
>> cp: cannot stat `/usr/local/genome/1Go.tzt': Permission denied
>>     
>
> That's expected.  You can't just copy the files across with scp, or
> rsync, or tar.
>
>   


I think, if you're _that_ low on budget, moving the whole scenario to a
vmware server setup is a good idea.
Marc's  http://www.open-sharedroot.org/  Website contains some
information about how to do this (although for a shared root cluster).

Are you preparing this for production?


cheers,
Rainer