In my experience, this kind of error could be related to write permission on the Share NFS.My suggestion is to check what user can read/write/execute your Share VolumeAll the bestDr. Alessio MoscatoMedical PhysicistMedical Physics Dep., Niguarda HospitalMilan, Italy2009/12/11 dti mic <[log in to unmask]>
Hi,I have set up the NFS between my machines A and B, and I can see the shared directory (/home/Share) in both machines. The NFS service seems to work correctly ( I run showmount, mount etc).Then, I put one data file into my shared directory (/home/Share), and run bedpostx again. This time, the job goes into error again, but the message is different:error: can't open output file "/home/Share/data/data.bedpostX/logsI tried to google it, but there is not much helpful info there. Your help is extremely important here.Bests!2009/12/10 wolf zinke <[log in to unmask]>
Hi,
In principle it is dead simple, once you found the information. I need to check whether i have somewhere the links to web pages describing it, but you might try to google for NIS and NFS (network information service and network file system).
You need to install the packages required for nfs (check your distributions details).
Lets call you PC's alpha and beta
To create the 'global' directory you just have to use mkdir on one of the PC's (e.g. on alpha: 'mkdir /GLOBAL_HD', this needs to be done as root) .
You have to make this directory available via the nfs service.
To do this you add a line to the '/etc/exports' file:
/GLOBAL_HD beta(rw,sync,no_wdelay,no_subtree_check,mp,no_root_squash)
Afterwards you need to restart the nfs service.
On the other PC (beta) you have to mount this directory by adding this line to the '/etc/fstab' file
alpha:/GLOBAL_HD /GLOBAL_HD nfs defaults 0 0
On beta you need the same directory as mount point as the the second entry in the added line says :
'mkdir /GLOBAL_HD'
with 'mount -a' you can activate the network directory.
I hope that this covers the steps and do not contain serious errors (I wrote it out of my memory at home, so I can't really guarantee that it is correct). Anyway, I would recommend to check your library for some good books covering this topic, once you managed it you realize how simple it is. However, to get there you might encounter several strange error messages ;-)
good luck,
wolf
On 12/10/2009 01:45 PM, micdtiserver wrote:Thank you very much for your suggestion! I am a beginner in SGE, so is there any detail instructions about setup the global directory between the hosts of my sge system? Or, is there any website I can look for? I am afraid I need some step by step help on this.2009-12-10
micdtiserver
发件人： wolf zinke发送时间： 2009-12-10 20:17:28收件人： FSL抄送：主题： Re: [FSL] Asking for your help about running bedpostx on SGEHi,Apparently you use the home directory for your data. In such a case youshould ensure that both computer use the same global home directory(that might be hardlinked to the local one) and have the same writingpermissions. For this purpose I use NIS. It also might be a good idea toset up a global directory for fsl and related software, so that you donot have several copies of it.Good luck anyway,wolfOn 12/10/2009 12:36 PM, Andrew Janke wrote:>> I set up SGE on two machines, namely A and B, each have 8 CPUs with platform>> CentOS 5 (AMD64). I set A as the qmaster, also as execd, and I set B as>> execd, too. The setup was done according to the guide on your website>> (ajanke), which goes well. After set up, I tried a test task>> (/sge/examples/simple.sh), which goes well, too.>>>> Then, I tried to run bedpostx for one sigle subject data. The data is>> uploaed to A, and submit. The job was distributed to the 8 CPUs of A>> immediately, however, when it is distributed to B, the job went Error. The>> message showed that: can't chdir to /home/<mydata>,>>> Have you set up B such that it NFS mounts user directories from A?> You should be able to log into either machine and see exactly the same> files.>>> Andrew>