clusternfs-users — ClusterNFS Users List

Re: [cNFS-Users] (no subject)

[Bernd S. <ber...@gm...>]

Hi Sandi,

On Friday 20 January 2006 13:13, sa...@hy... wrote:
> hi all, i'm trying to compile clusternfs but i'm getting this message:

don't know, havn't compiled clusternfs for a long time. I really suggest yo=
u=20
switch to unfs3 (unfs3.sf.net). Its better maintained, uses NFSv3 and has t=
he=20
same features as clusternfs.

Cheers,
	Bernd

=2D-=20
Bernd Schubert
PCI / Theoretische Chemie
Universit=E4t Heidelberg
INF 229
69120 Heidelberg

[cNFS-Users] (no subject)

[<sa...@hy...>]

hi all, i'm trying to compile clusternfs but i'm getting this message:



root@phoenix:/home/sandi/nfs/ClusterNFS_1.0.0-pre1# make
gcc -c  -DHAVE_CONFIG_H -DTRANSLATE_NAMES -DREADDIR_MASQUERADE -I. -I. -g=
 -O -
D_GNU_SOURCE -Wall -Wno-unused -Wno-switch -Wno-uninitialized -c mount_xd=
r.c
mount_xdr.c: In function `xdr_ppathcnf':
mount_xdr.c:221: warning: dereferencing `void *' pointer
mount_xdr.c:221: error: invalid use of void expression
mount_xdr.c:222: warning: dereferencing `void *' pointer
mount_xdr.c:222: error: invalid use of void expression
mount_xdr.c:223: warning: dereferencing `void *' pointer
mount_xdr.c:223: error: invalid use of void expression
mount_xdr.c:224: warning: dereferencing `void *' pointer
mount_xdr.c:224: error: invalid use of void expression
mount_xdr.c:225: warning: dereferencing `void *' pointer
mount_xdr.c:225: error: invalid use of void expression
mount_xdr.c:226: warning: dereferencing `void *' pointer
mount_xdr.c:226: error: invalid use of void expression
mount_xdr.c:245: warning: dereferencing `void *' pointer
mount_xdr.c:245: error: invalid use of void expression
mount_xdr.c:275: warning: dereferencing `void *' pointer
mount_xdr.c:275: error: void value not ignored as it ought to be
mount_xdr.c:276: warning: dereferencing `void *' pointer
mount_xdr.c:276: error: void value not ignored as it ought to be
mount_xdr.c:277: warning: dereferencing `void *' pointer
mount_xdr.c:277: error: void value not ignored as it ought to be
mount_xdr.c:278: warning: dereferencing `void *' pointer
mount_xdr.c:278: error: void value not ignored as it ought to be
mount_xdr.c:279: warning: dereferencing `void *' pointer
mount_xdr.c:279: error: void value not ignored as it ought to be
mount_xdr.c:280: warning: dereferencing `void *' pointer
mount_xdr.c:280: error: void value not ignored as it ought to be
mount_xdr.c:299: warning: dereferencing `void *' pointer
mount_xdr.c:299: error: void value not ignored as it ought to be
make: *** [mount_xdr.o] Error 1
root@phoenix:/home/sandi/nfs/ClusterNFS_1.0.0-pre1#=20

does anybody know how to fix it? please help.
i'm running on slackware 10.2
kernel is 2.6.12.4

thanks

[cNFS-Users] backbone of a large-file streaming system

[Nick I <clu...@gm...>]

Hi,
I help maintain a Web site at
www.clusterbuilder.org<http://www.clusterbuilder.org/>.
You might have seen before that I have a section called Ask the Cluster
Expert, where I am building a knowledgebase of cluster and grid information=
.
When someone asks a question I am researching the answer to build this
knowledgebase out. I received the following question:

*"I am looking for a variety of solutions to be a backbone of a large-file
streaming system providing thousands of concurrent download streams.
Preferably commodity hardware and Linux, though I'm open to commercial
solutions."*

I am wondering if anyone here has suggestions of what applications will wor=
k
best for this type of setup. You can respond to the question at
www.clusterbuilder.org/FAQ
<http://www.clusterbuilder.org/FAQ>or respond in email.

Thanks,
Nick

[cNFS-Users] TORQUE 2.0

[Amber W. <a_w...@ya...>]

Dear clusternfs-users,

I would like to announce that TORQUE Resource Manager
2.0 was just released, and can be downloaded at
www.clusterresources.com/torque.  

TORQUE, which is built on OpenPBS is one of the most
widely used open source batch schedulers.  TORQUE's
improvements since the last patch include an improved
start up feature for quick startup of downed nodes,
enhanced internal diagnostics, simplified install, and
improved API reporting abilities.

TORQUE is a community project with contributions from
NCSA, OSC, USC, the U.S. Department of Energy, Sandia,
PNNL, University of Buffalo, TeraGrid and many other
leading edge HPC organizations.

We invite you to download and try TORQUE and visit our
user community www.clusterresources.com/torque.  We
welcome feedback and patch submissions.

Thank you,

Amber 


__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com 

[cNFS-Users] Ask the Cluster Expert

[Nick I <clu...@gm...>]

Hi,

Thanks to the response from many in the community I have added sections
about diskless clusters and information on 32-bit and 64-bit processors at
the site I help run,
www.ClusterBuilder.org.<http://www.clusterbuilder.org/>I also added a
section called Ask the Cluster Expert (
http://www.clusterbuilder.org/pages/ask-the-expert.php) for people to submi=
t
questions they have about cluster and grid computing. I post the questions
at an FAQ page (http://www.clusterbuilder.org/pages/ask-the-expert/faq.php)
and then research the answer as well as allow those knowledgeable in the
community to submit a response to the question. I want to build a valuable
knowledgebase of high performance computing information.

I need you to share your knowledge by adding to the question responses and
also submitting questions/answers to common problems you've experienced in
the past and are experiencing now.

Thanks,

Nick

Re: [cNFS-Users] Hiding files to clients

[Bernd S. <ber...@we...>]

On Saturday 09 August 2003 10:44, axlmac\@inwind\.it wrote:
> Hi all clusterNFS users,
>   I decide to use cluserNFS for my purposes in creating a cluster of Pc for
> parallel programming. I note cNFS is more flexible and manageable than the
> classic nfs daemon. By default cNFS shares all files.I think it is problem
> when OS (like linux) uses files as flags to remember it that something
> happens or not.
> In practice:
> I shared the entire root partion of the server to the clients (I have
> omogeneous PCs). Each time a client starts, it looks for /.autofsck (the
> file exists between the mount action of this filesystems and correspect
> umount of the same one).
> If it finds out it a force check is performed on that filesystem.
> The problem is that file is shared about all server and clients!
> The result is clients performe check everytime they start because server
> created it. How (and this is, finally, my question) can I hide files to
> clients?
>
> Have I change the behaviour of master node forcing it to manage a
> /.autofsck$$HOST=MASTER$$ instead of /.autofsck so client never will see
> it?
> This case I must controle all OS script where /.autofsck is mentioned and I
> would not correct files of my distrbution (RH 8.0)
>
> Did you encounter the same problem? How did you resolv it?
>
> Thanks to everybody who will help me.
>
> Alex Macuz
>

Hello Alex,

I suggest to use the cvs-version, since the behaviour of the $$CREATE=IP$$ tag 
has changed there (actually a side effect of making the CREATE-tag more 
reliable). Just an example:

/etc/test
/etc/test$$CREATE=IP$$

Using the non-cvs-Version, the clients still would see /etc/test, using the 
cvs-version, the file simply doesn't exist for the clients, unless there are 
client specific files as /etc/test$$IP=x.y.z$$.

Also I don't know how much you depend on high availability. In our group all 
workstations are used for usual work (programming, writing, etc.) but also 
for our calculations. So we can't reboot very often, in fact we only reboot 
one or two times a year. Using only ClusterNFS for this pupose wasn't 
reliable enough and furthermore rather slow. So we are exporting the hole 
root-fs by knfsd and additionally only /etc and /var by ClusterNFS. This has 
proven to be very very stable and also much faster than pure CNFSD, but also 
requires some knowledge about initrd's and nfs- and rpc-ports.
I'm planning to write a HOWTO about our Cluster, but so far didn't have the 
time to do so. Just tell me, if you would like to know deeper insights.

Hope it helps,
	Bernd

[cNFS-Users] Hiding files to clients

[axlmac\@inwind\.it <ax...@in...>]

Hi all clusterNFS users,
  I decide to use cluserNFS for my purposes in =
creating a cluster of Pc for parallel programming.
I note cNFS is more f=
lexible and manageable than the classic nfs daemon.
By default cNFS shar=
es all files.I think it is problem when OS (like linux) uses files as fla=
gs 
to remember it that something happens or not.
In practice:
I share=
d the entire root partion of the server to the clients (I have omogeneous=
 PCs).
Each time a client starts, it looks for /.autofsck (the file exis=
ts between the mount action 
of this filesystems and correspect umount o=
f the same one).
If it finds out it a force check is performed on that f=
ilesystem.
The problem is that file is shared about all server and clien=
ts!
The result is clients performe check everytime they start because se=
rver created it.
How (and this is, finally, my question) can I hide file=
s to clients?

Have I change the behaviour of master node forcing it to=
 manage a /.autofsck$$HOST=3DMASTER$$ instead of /.autofsck 
so client n=
ever will see it?
This case I must controle all OS script where /.autofs=
ck is mentioned and I would not correct files of my 
distrbution (RH 8.0=
)

Did you encounter the same problem? How did you resolv it?

Thanks=
 to everybody who will help me.

Alex Macuz

Re: [cNFS-Users] using dhcpd

[Michal L. <mic...@lo...>]

On Mon, 30 Jun 2003, Philippe Amelant wrote:

> Le lun 30/06/2003 ? 14:59, Michal Ludvig a =E9crit :
> > BTW Why do you insist on bootp? Most recent DHCP servers can respond to
> > bootp queries as well...
>
> hello
> I don't want to use MAC addresses for my cluster so I need dhcp instead
> bootp (all doc say that i need all mac addresses I want to put in the
> cluster). It work well with dhcp and a range of IP addresses
> thank you for your help

I'm not sure if I understand you correctly...

Are you saying that you don't know all MAC addresses of all nodes in the
cluster? That's not a problem - just instruct DHCP to use a range of
addresses and it will automatically distribute them over the nodes
regardless if the nodes ask via DHCP or BOOTP protocol. Usually DHCP
servers are so intelligent that they remember which address they gave to
which client and don't change it unless necesarry.

Antoher option is to use nodes with harddisks and boot from them - this
way you could configure the IP address on each node locally.

I have a feeling that you could use DHCP and the problems you are afraid
of come from a kind misunderstanding only. But I may be wrong ;-)

Anyway if you could come up with more details I'll try to provide further
help.

Michal Ludvig
--=20
A: No.
Q: Should I include quotations after my reply?

Re: [cNFS-Users] using dhcpd

[Philippe A. <phi...@fr...>]

Le lun 30/06/2003 à 14:59, Michal Ludvig a écrit :
> On Mon, 16 Jun 2003, Philippe Amelant wrote:
> 
> > I am testing to make a small cluster with open mosix and clusternfs
> > Is it possible to use dhcp instead of bootpd ?
> > (I haven't found option in dhcpd doc for root=... option)
> 
> Should be. It depends on your kernel configuration. There is an option in
> the kernel config where you could choose if it should try bootp, dhcp or
> even rarp (IIRC). Check the kernel doco for details.
> It's also a question if your network cards can boot with dhcp instead
> of bootp.
> 
> BTW Why do you insist on bootp? Most recent DHCP servers can respond to
> bootp queries as well...

hello
I don't want to use MAC addresses for my cluster so I need dhcp instead
bootp (all doc say that i need all mac addresses I want to put in the
cluster). It work well with dhcp and a range of IP addresses
thank you for your help 


> 
> Michal Ludvig
-- 
Philippe Amelant <phi...@fr...>

Re: [cNFS-Users] using dhcpd

[Michal L. <mi...@lo...>]

On Mon, 16 Jun 2003, Philippe Amelant wrote:

> I am testing to make a small cluster with open mosix and clusternfs
> Is it possible to use dhcp instead of bootpd ?
> (I haven't found option in dhcpd doc for root=... option)

Should be. It depends on your kernel configuration. There is an option in
the kernel config where you could choose if it should try bootp, dhcp or
even rarp (IIRC). Check the kernel doco for details.
It's also a question if your network cards can boot with dhcp instead
of bootp.

BTW Why do you insist on bootp? Most recent DHCP servers can respond to
bootp queries as well...

Michal Ludvig
-- 
A: No.
Q: Should I include quotations after my reply?

[cNFS-Users] using dhcpd

[Philippe A. <phi...@fr...>]

Hello 

I am testing to make a small cluster with open mosix and clusternfs
Is it possible to use dhcp instead of bootpd ?
(I haven't found option in dhcpd doc for root=... option)

bye
thank 


-- 
Philippe Amelant <phi...@fr...>

Re: [cNFS-Users] Question from a newbie

[Bernd S. <ber...@we...>]

On Monday 09 June 2003 16:19, Warnes, Gregory R wrote:
> Most unix shells convert the sequence $$ to the process ID of the shell. 
> To create a file with these characters from the shell, you need to excape
> the dollar sign characters, like so:
>
> 	touch filename\$\$CLIENT\$\$
>
> The double dollar-sign character was explicitly chosen so that it would be
> very unlikely to appear in any normal file name...
>
> -Greg

Hi,

another way is to use '...', so: touch 'filename$$CLIENT$$'

Bernd

RE: [cNFS-Users] Question from a newbie

[Warnes, G. R <gre...@gr...>]

Most unix shells convert the sequence $$ to the process ID of the shell.  To
create a file with these characters from the shell, you need to excape the
dollar sign characters, like so:

	touch filename\$\$CLIENT\$\$

The double dollar-sign character was explicitly chosen so that it would be
very unlikely to appear in any normal file name...

-Greg



> -----Original Message-----
> From: Michael Langan [mailto:Mic...@gm...]
> Sent: Monday, June 09, 2003 10:09 AM
> To: clu...@li...
> Subject: [cNFS-Users] Question from a newbie
> 
> 
> Hello,
> 
> I'm new to MOSIX / ClusterNFS configuration. I'm following 
> the Queenbee
> recipe at the moment but am stuck when trying to create the $$CLIENT$$
> files. Every time I create them, instead of getting the '$$' 
> bit I get a
> spurious number instead e.g. 1092CLIENT1092 etc. (Doing an 'echo $$'
> will also return the same number.) The shell I'm using is 
> /bin/bash. Is
> there any way around this? Is it critical? Does anybody know why this
> should happen?
> 
> Reply personnally if you like.
> Many TIA, 
> Mike.
> 
> 
> 
> -------------------------------------------------------
> This SF.net email is sponsored by:  Etnus, makers of 
> TotalView, The best
> thread debugger on the planet. Designed with thread debugging features
> you've never dreamed of, try TotalView 6 free at www.etnus.com.
> _______________________________________________
> ClusterNFS-Users mailing list
> Clu...@li...
> https://lists.sourceforge.net/lists/listinfo/clusternfs-users
> 


LEGAL NOTICE
Unless expressly stated otherwise, this message is confidential and may be privileged. It is intended for the addressee(s) only. Access to this E-mail by anyone else is unauthorized. If you are not an addressee, any disclosure or copying of the contents of this E-mail or any action taken (or not taken) in reliance on it is unauthorized and may be unlawful. If you are not an addressee, please inform the sender immediately.

[cNFS-Users] Question from a newbie

[Michael L. <Mic...@gm...>]

Hello,

I'm new to MOSIX / ClusterNFS configuration. I'm following the Queenbee
recipe at the moment but am stuck when trying to create the $$CLIENT$$
files. Every time I create them, instead of getting the '$$' bit I get a
spurious number instead e.g. 1092CLIENT1092 etc. (Doing an 'echo $$'
will also return the same number.) The shell I'm using is /bin/bash. Is
there any way around this? Is it critical? Does anybody know why this
should happen?

Reply personnally if you like.
Many TIA, 
Mike.

[cNFS-Users] Re: [NFS] knfsd & unfsd on the same server

[Bernd S. <ber...@we...>]

On Thursday 27 February 2003 18:54, Trond Myklebust wrote:
> >>>>> " " == Bernd Schubert <ber...@we...> writes:
>      > mount -t nfs -o mountprog=110005,nfsprog=110003,port=2050
>      > localhost:/test1 /mnt/test2
>
> The 'nfsprog' option is unsupported by the Linux NFS client as the
> mount program has no way of passing the number down to the kernel.
>
> Cheers,
>   Trond

Thanks for your answer, so we know that we will have to live with this issue.

Best regards,
	Bernd

[cNFS-Users] Re: [NFS] knfsd & unfsd on the same server

[Trond M. <tro...@fy...>]

>>>>> " " == Bernd Schubert <ber...@we...> writes:

     > mount -t nfs -o mountprog=110005,nfsprog=110003,port=2050
     > localhost:/test1 /mnt/test2

The 'nfsprog' option is unsupported by the Linux NFS client as the
mount program has no way of passing the number down to the kernel.

Cheers,
  Trond

[cNFS-Users] knfsd & unfsd on the same server

[Bernd S. <ber...@we...>]

Hi,

I would like to get knfsd and unfsd (actually ClusterNFS) running on the same 
server. Actually this already working, even without modifying the source 
code, when e.g. unfsd is running on another port. 
However, on stopping any of both servers, both are unregistered from rpc, as 
rpcinfo shows. Furthermore there are strange log-file messages on mounting 
from unfsd, probably since first the wrong rpc.mountd is being called.

To improve this I tried to change the rpc-program numbers in unfsd. For the  
rpc.mountd from unfsd this works fine, however when I *also* change the 
rpc-number for the nfsd (unfsd), I can't mount from this server. 
Here are 2 example commands:

mount -t nfs localhost:/test1	/mnt/test1 #would mount from knfsd

mount -t nfs -o mountprog=110005,nfsprog=110003,port=2050 localhost:/test1 
/mnt/test2


So again, when I only change the mountd-rpc-program-number, the second mount 
command would work (of course without specifying nfsprog=...), even without 
any error/warning message in the logfiles. However, this still causes that 
both nfsd-servers are unregistered from rpc when any of both is being stopped

Is this a bug in the mount command, or did I make a mistake when I changed the 
rpc-program numbers in unfsd? (I changed them in nfs_prot.h and mount.h). At 
least 'rpcinfo -p localhost' seems to show me that unfsd is working with the 
changed rpc-number.


Thanks in advance for any help,
	Bernd

Re: [cNFS-Users] Debian packages?

[Michal L. <mi...@lo...>]

On Mon, 27 Jan 2003, John Buttery wrote:

>   I was just wondering if there were any plans to create new .deb
> packages for the 3.x versions...I see that there are .deb packages for
> the older version.  Any chance?  :) 

Sure. If you do them I'll upload'em to sf.net :-)

Michal Ludvig

[cNFS-Users] Debian packages?

[John B. <jo...@io...>]

  I was just wondering if there were any plans to create new .deb
packages for the 3.x versions...I see that there are .deb packages for
the older version.  Any chance?  :)=20

--=20
------------------------------------------------------------------------
 John Buttery
                                     (Web page temporarily unavailable)
------------------------------------------------------------------------

RE: [cNFS-Users] swapping over nfs

[Warnes, G. R <gre...@gr...>]

> -----Original Message-----
> From: Rick Rauenzahn [mailto:ri...@la...]
> 
> it seems like clusternfs requires that swapping, if any, is done
> over nfs. what are your experiences with this?  we have large
> memory nodes, but potentially large memory jobs as well.  is
> there any way to define a local swapping device?  we have the
> disk space available, but i like clusternfs management of the
> root filesystem.

You can set up swapping however you want.  To use a local swap partition,
just set it up in the approprate 
 /etc/fstab$$clients$$
(or whatever) file.   The last cluster I set up using ClusterNFS operated in
just this manner.

-Greg

> 
> thanks.
> 
> Rick
> 
> 
> -------------------------------------------------------
> This sf.net email is sponsored by:ThinkGeek
> Welcome to geek heaven.
> http://thinkgeek.com/sf
> _______________________________________________
> ClusterNFS-Users mailing list
> Clu...@li...
> https://lists.sourceforge.net/lists/listinfo/clusternfs-users
> 


LEGAL NOTICE
Unless expressly stated otherwise, this message is confidential and may be privileged. It is intended for the addressee(s) only. Access to this E-mail by anyone else is unauthorized. If you are not an addressee, any disclosure or copying of the contents of this E-mail or any action taken (or not taken) in reliance on it is unauthorized and may be unlawful. If you are not an addressee, please inform the sender immediately.

[cNFS-Users] swapping over nfs

[Rick R. <ri...@la...>]

it seems like clusternfs requires that swapping, if any, is done
over nfs.  what are your experiences with this?  we have large
memory nodes, but potentially large memory jobs as well.  is
there any way to define a local swapping device?  we have the
disk space available, but i like clusternfs management of the
root filesystem.

thanks.

Rick

[cNFS-Users] Re: [Ltsp-discuss] Using ClusterNFS to enable thin clients to use local cdrom drive

[Jason A. P. <pat...@pc...>]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

First of all, you will need to setup NFS or some equivalent on each
thin-client in order to export the cdrom resource of the thin-client.
You might take a look at ENBD which has enhancements that have been
added in order to export remotely floppy and cdrom resources from
machines (specifically for the thin-client problem) and have them be
able to be accessed on the application server.

Brent Hasty wrote:
> I am using ClusterNFS to enable thin clients to use local cdrom drives.
> Working with Mandrake Linux 9.0 to set up the thin client terminals, and
> setting up files and directories per client as per the clusternfs way.
> cdrom$$hostname=workstation1$$ however the workstation is attempting
to acess
> the cdrom on the server, not the cdrom locally installed in the terminals
> computer case.
>
> Is there documentation specific to thin clients avaliable??
> Can a thin client acess its cdrom n floppy via clusternfs, while all
apps are
> running on the server using xdmcmp?
>
>
> -------------------------------------------------------
> This sf.net email is sponsored by: See the NEW Palm
> Tungsten T handheld. Power & Color in a compact size!
> http://ads.sourceforge.net/cgi-bin/redirect.pl?palm0001en
> _____________________________________________________________________
> Ltsp-discuss mailing list.   To un-subscribe, or change prefs, goto:
>       https://lists.sourceforge.net/lists/listinfo/ltsp-discuss
> For additional LTSP help,   try #ltsp channel on irc.openprojects.net
>

- --
Jason A. Pattie
pat...@pc...
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.0.6 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQE90CcuuYsUrHkpYtARAlerAJ9TLr5JOL37yekw26hUoYIsI5UV+ACdElNi
hkz5O6HdUjYHa5Xh3A2F9Ok=
=Yojy
-----END PGP SIGNATURE-----


-- 
This message has been scanned for viruses and
dangerous content by MailScanner, and is
believed to be clean.

Re: [cNFS-Users] Using ClusterNFS to enable thin clients to use local cdrom drive

[Michal L. <mi...@lo...>]

On Tue, 5 Nov 2002, Brent Hasty wrote:

> Can a thin client acess its cdrom n floppy via clusternfs, while all apps are 
> running on the server using xdmcmp?

You'd have to export terminals' CDROMs via NFS and mount them on the 
server to appropriate directories. Then it might be possible. However I'm 
not sure how to mount and unmount disks in those drives...

Michal Ludvig

[cNFS-Users] Using ClusterNFS to enable thin clients to use local cdrom drive

[Brent H. <blu...@cr...>]

I am using ClusterNFS to enable thin clients to use local cdrom drives. =20
Working with Mandrake Linux 9.0 to set up the thin client terminals, and=20
setting up files and directories per client as per the clusternfs way. =20
cdrom$$hostname=3Dworkstation1$$ however the workstation is attempting to=
 acess=20
the cdrom on the server, not the cdrom locally installed in the terminals=
=20
computer case.

Is there documentation specific to thin clients avaliable??
Can a thin client acess its cdrom n floppy via clusternfs, while all apps=
 are=20
running on the server using xdmcmp?