[Pvfs2-users] PVFS2 Installation
jpummil at uark.edu
Thu Apr 20 06:27:03 EDT 2006
Things are looking pretty good on the installation!
I think the oddity with the pvfs2-rm command lies in the
/mnt/pvfs2/sample structure. If I just specify /mnt/pvfs2/<filename>,
the pvfs2-rm will delete the file with no complaints...even though the
file appears to be nested within the /sample subdirectory. Another
Regarding the compute node clients, I think the "roll" set them up and
*[jpummil at prospero ~]$ ssh c0-0
Last login: Tue Apr 18 09:06:14 2006 from prospero.local
Rocks Compute Node
Rocks 4.0.0 (Whitney)
Profile built 14:11 12-Apr-2006
Kickstarted 14:17 12-Apr-2006
[jpummil at compute-0-0 ~]$ ps -ef | grep pvf
root 7321 1 0 Apr12 ? 00:00:20 /usr/sbin/automount
--timeout=60 --ghost /mnt/pvfs2 file /etc/auto.pvfs2
root 7370 1 0 Apr12 ? 00:00:00
/opt/pvfs2/sbin/pvfs2-client -p /opt/pvfs2/sbin/pvfs2-client-core
root 7371 7370 0 Apr12 ? 00:00:00 pvfs2-client-core -a 5
jpummil 9267 9244 0 10:20 pts/0 00:00:00 grep pvf
What is the significance of the commands in the /pvfs2/bin directory?
For example, pvfs2-cp vs cp and pvfs2-rm vs rm? Does this have anything
to do with whether or not the kernel module is running (someone told me
that if the kernel module for pvfs2 was running properly, you could use
the standard linux commands interchangeably...)?
Jeff F. Pummill
Robert Latham wrote:
> On Wed, Apr 19, 2006 at 05:52:40AM -0500, Jeff Pummill wrote:
>> Morning Rob,
>> The -t option was just what I needed! Thanks!
>> Interestingly enough, when I try and delete the file with pvfs2-rm, I
>> get an error...
>> [root at compute-0-0 sample]# pwd
>> [root at compute-0-0 sample]# ls
>> bio2.iso biobrew-4.1-0.i386.iso libgfortran-4.0.2-8.fc4.i386.rpm
>> [root at compute-0-0 sample]# /opt/pvfs2/bin/pvfs2-rm libgfortran-4.0.2-8.fc4.i386.rpm
>> PVFS_util_resolve: No such file or directory
> pvfs2-rm does need the full pvfs2 path to the pvfs2 file, so this
> failure is expected (but perhaps poorly documented)
>> [root at compute-0-0 sample]# /opt/pvfs2/bin/pvfs2-rm /mnt/pvfs2/sample/libgfortran-4.0.2-8.fc4.i386.rpm
>> PVFS_sys_lookup: No such file or directory
>> The file I am trying to delete was copied with pvfs2-cp to its present
>> location. Am I doing something incorrectly?
> That way looks right to me. Does 'pvfs2-ping -m /mnt/pvfs2' still
> look ok? (it's possible a server for any number of reasons crashed
>> Also, as I have 24 compute nodes who have been assigned different I/O
>> servers (4 compute nodes per server), what happenss if an I/O server
>> goes down? Will those 4 compute nodes be unable to access the pvfs2
> Any compute node accessing PVFS2 will by default stripe data across
> all pvfs2 servers. The assignment you made in the fstab was just
> telling the client which server to contact first.
> If any pvfs2 server goes down, you will be unable to access pvfs2
> files until that server comes back up. Don't put the only copy of
> important data on pvfs2: it's meant to be a fast scratch space, so
> keep backups.
>> Regarding the kernel...
>> [root at prospero ~]# lsmod
>> Module Size Used by
>> nfsd 205153 9
>> exportfs 10049 1 nfsd
>> lockd 65129 2 nfsd
>> md5 8001 1
>> ipv6 238945 27
>> pvfs2 43396 1
>> ...this shows we're good...right?
> It's a very good start. Two steps remain:
> - Start pvfs2-client and pvfs2-client-core
> - Mount the file system.
> I can't give you the exact commands for these steps since they are so
> dependent on how you have installed and configured pvfs2, but they
> will look something like this (adjust the paths and server names for
> your environment):
> prompt# /opt/pvfs2/sbin/pvfs2-client -p /opt/pvfs2/sbin/pvfs2-client-core
> prompt# mount -t pvfs2 tcp://io-node:3334/pvfs2-fs /mnt/pvfs2
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Pvfs2-users