bproc-users — General discussion about BProc.

Re: [BProc] BProc with mixed architecture

[Erik H. <er...@he...>]

On 9/22/05, Jeff Rasmussen <jra...@ln...> wrote:
> Does anyone know of a way to run bproc in a mixed 32/64 bit
> architecture?
>
> I need a 64bit host with 64bit kernel to be able to boot both 64bit and
> 32bit nodes. Anyone know if this is possible with the current software
> stack?

It's not... and for good reason.  If you look under the covers
(VMADump process layout and message packing details), x86 and x86_64
are almost as different as x86 and ppc64.  The register files are
different, memory space details are different, etc.  A 32 bit process
dumped on x86_64 is not directly undumpable on x86 and vice versa.

- Erik

[BProc] BProc with mixed architecture

[Jeff R. <jra...@ln...>]

Does anyone know of a way to run bproc in a mixed 32/64 bit
architecture? 

I need a 64bit host with 64bit kernel to be able to boot both 64bit and
32bit nodes. Anyone know if this is possible with the current software
stack?

Thanks,

-- 
Jeff Rasmussen

[BProc] Re: Re: mpif77 and mpif90 in Clustermatic 5

[Tomasz P. <bp...@o2...>]

Hello,

> Daniel Gruner wrote:
> Actually, all you need to do is install the package mpich-1.2.5.2-3, fo=
r
> which the .src.rpm is included in the cm5 distribution.  Just recompile
> the mpich-1.2.5.2-3.src.rpm on your system, and install it.
>=20
> The only trick afterwards is that you MUST use the /usr/bin/mpirun that
> is part of the cmtools package, which seems not to be in the cm5 distri=
bution.
> cmtools-1.4 is in the sourceforge.net/projects/bproc page, and you can
> get it from there and build it for your system.

I finally rebuild the mpich-1.2.5.3.src.rpm (delivered with CM5) and=20
install it (of course, before I removed standard CM5 version), but still=20
I don't have any fortran compiler in /usr/mpich-p4/bin.

Any suggestions what can I do to compile parallel fortran program?

Thank you in advance for adivices.

--=20

Tomasz Perlik

V FDK
Faculty of Physics, Astronomy and Informatics
Nicolaus Copernicus University, Toru=F1, POLAND

[BProc] anyone refer beowulf documentation

[kumanan c. <ku...@gm...>]

can anyone refer me beowulf documentation so that it would helpful for me t=
o=20
implement in linux
basic as well as indepth documentation

Re: [BProc] run processes on master and modify machines file

[Greg W. <gw...@la...>]

The machines file is not used. Instead, mpirun will allocate nodes  
that are available to you for execution (as shown by bpstat) or nodes  
that you specify using the NODES environment variable.

Greg

On Sep 1, 2005, at 4:53 PM, Andrew Pitre wrote:

> I've setup bproc on my development cluster (1 master, 1 slave).
> Commands execute successfully on the slave using bpsh.
>
> The clustermatic 5 version of mpich seems to be installed and working,
> however, i don't seem to be able to alter the machines file that  
> mpich uses
> to allocate processes to nodes.
>
> I've tried editing:
> /usr/mpich-p4/share/machines.LINUX
> and
> /util/machines/machines.LINUX
> to contain the ip addresses of my two nodes:
> 10.0.1.100
> 10.0.1.101
>
> > mpirun -np 2 mpiTest
> Not enough nodes to allocate all processes
>
> alternately
> > mpirun -np 1 mpiTest
> Test successful on process number: 0
> elapsed_time =
>         8.3e-05
>
> Can anyone tell me how to edit the machines file, where it is  
> located, and how to run on the master node.  Can mpich do round  
> robin process allocation within this machines file?
>
> - Andrew
>
>
> -------------------------------------------------------
> SF.Net email is Sponsored by the Better Software Conference & EXPO
> September 19-22, 2005 * San Francisco, CA * Development Lifecycle  
> Practices
> Agile & Plan-Driven Development * Managing Projects & Teams *  
> Testing & QA
> Security * Process Improvement & Measurement * http://www.sqe.com/ 
> bsce5sf
> _______________________________________________
> BProc-users mailing list
> BPr...@li...
> https://lists.sourceforge.net/lists/listinfo/bproc-users
>

[BProc] run processes on master and modify machines file

[Andrew P. <ap...@ro...>]

I've setup bproc on my development cluster (1 master, 1 slave).
Commands execute successfully on the slave using bpsh.

The clustermatic 5 version of mpich seems to be installed and working,
however, i don't seem to be able to alter the machines file that  
mpich uses
to allocate processes to nodes.

I've tried editing:
/usr/mpich-p4/share/machines.LINUX
and
/util/machines/machines.LINUX
to contain the ip addresses of my two nodes:
10.0.1.100
10.0.1.101

 > mpirun -np 2 mpiTest
Not enough nodes to allocate all processes

alternately
 > mpirun -np 1 mpiTest
Test successful on process number: 0
elapsed_time =
         8.3e-05

Can anyone tell me how to edit the machines file, where it is  
located, and how to run on the master node.  Can mpich do round robin  
process allocation within this machines file?

- Andrew

[BProc] v9fs and bproc

[Daniel G. <dg...@cp...>]

Hi

Is anybody using v9fs file systems with bproc?  If so, how
well does it work?  I am looking for an alternative to NFS
which I suspect of causing crashes to my master node
(Dual Opteron, running Clustermatic 5 on top of Fedora Core 2).

Daniel
-- 

Dr. Daniel Gruner                        dg...@ch...
Dept. of Chemistry                       dan...@ut...
University of Toronto                    phone:  (416)-978-8689
80 St. George Street                     fax:    (416)-978-5325
Toronto, ON  M5S 3H6, Canada             finger for PGP public key

Re: [BProc] Reply to : mpif77 and mpif90 in Clustermatic 5 (Tomasz Perlik

[Daryl W. G. <dw...@la...>]

> Date: Tue, 30 Aug 2005 12:13:42 -0400
> From: Daniel Gruner <dg...@cp...>
> To: ro...@or...
> Cc: BProc users list <bpr...@li...>
> Subject: Re: [BProc] Reply to : mpif77 and mpif90 in Clustermatic 5 (Tomasz Perlik
> 
> Actually, all you need to do is install the package mpich-1.2.5.2-3, for
> which the .src.rpm is included in the cm5 distribution.  Just recompile
> the mpich-1.2.5.2-3.src.rpm on your system, and install it.

FWIW, LA-MPI (http://public.lanl.gov/lampi/) also contains support for
BProc, but again you must use their mpirun to launch your job.  This code
is being deprecated, however, as the LA-MPI guys have joined forces with
LAM (and others) to create Open MPI (http://www.open-mpi.org/) which,
presumably, will also have BProc support from the get-go.

Daryl

> 
> The only trick afterwards is that you MUST use the /usr/bin/mpirun that
> is part of the cmtools package, which seems not to be in the cm5 distribution.
> cmtools-1.4 is in the sourceforge.net/projects/bproc page, and you can
> get it from there and build it for your system.
> 
> If you need the cmtools source package let me know and I can send it to
> you.
> 
> Regards,
> Daniel
> 
> On Tue, Aug 30, 2005 at 05:15:19PM +0100, ro...@or... wrote:
> > 
> > I have the same problem. I use Clustermatic 5 for our cluster and some programs
> > ( MPQC ) requires a fortran compiler. Clustermatic5 does not have any so I took
> > another mpich that had all the compilers present (mpich-p4, I don't remember
> > from where, but it should not be hard to find via google). When I e-mailed the
> > staff at clustermatic (www.clustermatic.org) they said that this could be a
> > possible approach, however one has to change the libmpich.a to the one supplied
> > with the clustermatic5. But if this does not work one has to e-mail them,
> > becuase one then has to try the second approach, which involves recompiling the
> > whole thing from the begining. This has to be done to get the compilers working
> > for bproc type systems.
> > I hope this can be of some help to you.
> > 
> > /Sincerely
> > Robert Eklund, Ph D
> > 
> > 
> > Quoting bpr...@li...:
> > 
> > > Send BProc-users mailing list submissions to
> > > 	bpr...@li...
> > >
> > > To subscribe or unsubscribe via the World Wide Web, visit
> > > 	https://lists.sourceforge.net/lists/listinfo/bproc-users
> > > or, via email, send a message with subject or body 'help' to
> > > 	bpr...@li...
> > >
> > > You can reach the person managing the list at
> > > 	bpr...@li...
> > >
> > > When replying, please edit your Subject line so it is more specific
> > > than "Re: Contents of BProc-users digest..."
> > >
> > >
> > > Today's Topics:
> > >
> > >    1. mpif77 and mpif90 in Clustermatic 5 (Tomasz Perlik)
> > >
> > > -- __--__-- 
> > >
> > > Message: 1
> > > Date: Tue, 30 Aug 2005 02:57:36 -0400
> > > From: Tomasz Perlik <bp...@o2...>
> > > To: bpr...@li...
> > > Subject: [BProc] mpif77 and mpif90 in Clustermatic 5
> > >
> > > Helo,
> > >
> > > I try compile my program in fortran on Clustermatic 5 and I don't have=20
> > > compilers such mpif77 and mpif90? I remember that in Clustermatic 4 I=20
> > > used mpif77. Why? Maybe I do something wrong or there is another way?
> > >
> > > Any suggestion is appreciated.
> > >
> > > --=20
> > >
> > > Tomasz Perlik
> > >
> > > V FDK
> > > Faculty of Physics, Astronomy and Informatics
> > > Nicolaus Copernicus University, Toru=F1, POLAND
> > >
> > >
> > >
> > >
> > >
> > > -- __--__-- 
> > >
> > > _______________________________________________
> > > BProc-users mailing list
> > > BPr...@li...
> > > https://lists.sourceforge.net/lists/listinfo/bproc-users
> > >
> > >
> > > End of BProc-users Digest
> > >
> > 
> > 
> > --
> > 
> > 
> > 
> > 
> > -------------------------------------------------------
> > SF.Net email is Sponsored by the Better Software Conference & EXPO
> > September 19-22, 2005 * San Francisco, CA * Development Lifecycle Practices
> > Agile & Plan-Driven Development * Managing Projects & Teams * Testing & QA
> > Security * Process Improvement & Measurement * http://www.sqe.com/bsce5sf
> > _______________________________________________
> > BProc-users mailing list
> > BPr...@li...
> > https://lists.sourceforge.net/lists/listinfo/bproc-users
> 
> -- 
> 
> Dr. Daniel Gruner                        dg...@ch...
> Dept. of Chemistry                       dan...@ut...
> University of Toronto                    phone:  (416)-978-8689
> 80 St. George Street                     fax:    (416)-978-5325
> Toronto, ON  M5S 3H6, Canada             finger for PGP public key
> 
> 
> 

-- 
			***** Correspondence *****

Re: [BProc] Reply to : mpif77 and mpif90 in Clustermatic 5 (Tomasz Perlik

[Daniel G. <dg...@cp...>]

Actually, all you need to do is install the package mpich-1.2.5.2-3, for
which the .src.rpm is included in the cm5 distribution.  Just recompile
the mpich-1.2.5.2-3.src.rpm on your system, and install it.

The only trick afterwards is that you MUST use the /usr/bin/mpirun that
is part of the cmtools package, which seems not to be in the cm5 distribution.
cmtools-1.4 is in the sourceforge.net/projects/bproc page, and you can
get it from there and build it for your system.

If you need the cmtools source package let me know and I can send it to
you.

Regards,
Daniel

On Tue, Aug 30, 2005 at 05:15:19PM +0100, ro...@or... wrote:
> 
> I have the same problem. I use Clustermatic 5 for our cluster and some programs
> ( MPQC ) requires a fortran compiler. Clustermatic5 does not have any so I took
> another mpich that had all the compilers present (mpich-p4, I don't remember
> from where, but it should not be hard to find via google). When I e-mailed the
> staff at clustermatic (www.clustermatic.org) they said that this could be a
> possible approach, however one has to change the libmpich.a to the one supplied
> with the clustermatic5. But if this does not work one has to e-mail them,
> becuase one then has to try the second approach, which involves recompiling the
> whole thing from the begining. This has to be done to get the compilers working
> for bproc type systems.
> I hope this can be of some help to you.
> 
> /Sincerely
> Robert Eklund, Ph D
> 
> 
> Quoting bpr...@li...:
> 
> > Send BProc-users mailing list submissions to
> > 	bpr...@li...
> >
> > To subscribe or unsubscribe via the World Wide Web, visit
> > 	https://lists.sourceforge.net/lists/listinfo/bproc-users
> > or, via email, send a message with subject or body 'help' to
> > 	bpr...@li...
> >
> > You can reach the person managing the list at
> > 	bpr...@li...
> >
> > When replying, please edit your Subject line so it is more specific
> > than "Re: Contents of BProc-users digest..."
> >
> >
> > Today's Topics:
> >
> >    1. mpif77 and mpif90 in Clustermatic 5 (Tomasz Perlik)
> >
> > --__--__--
> >
> > Message: 1
> > Date: Tue, 30 Aug 2005 02:57:36 -0400
> > From: Tomasz Perlik <bp...@o2...>
> > To: bpr...@li...
> > Subject: [BProc] mpif77 and mpif90 in Clustermatic 5
> >
> > Helo,
> >
> > I try compile my program in fortran on Clustermatic 5 and I don't have=20
> > compilers such mpif77 and mpif90? I remember that in Clustermatic 4 I=20
> > used mpif77. Why? Maybe I do something wrong or there is another way?
> >
> > Any suggestion is appreciated.
> >
> > --=20
> >
> > Tomasz Perlik
> >
> > V FDK
> > Faculty of Physics, Astronomy and Informatics
> > Nicolaus Copernicus University, Toru=F1, POLAND
> >
> >
> >
> >
> >
> > --__--__--
> >
> > _______________________________________________
> > BProc-users mailing list
> > BPr...@li...
> > https://lists.sourceforge.net/lists/listinfo/bproc-users
> >
> >
> > End of BProc-users Digest
> >
> 
> 
> --
> 
> 
> 
> 
> -------------------------------------------------------
> SF.Net email is Sponsored by the Better Software Conference & EXPO
> September 19-22, 2005 * San Francisco, CA * Development Lifecycle Practices
> Agile & Plan-Driven Development * Managing Projects & Teams * Testing & QA
> Security * Process Improvement & Measurement * http://www.sqe.com/bsce5sf
> _______________________________________________
> BProc-users mailing list
> BPr...@li...
> https://lists.sourceforge.net/lists/listinfo/bproc-users

-- 

Dr. Daniel Gruner                        dg...@ch...
Dept. of Chemistry                       dan...@ut...
University of Toronto                    phone:  (416)-978-8689
80 St. George Street                     fax:    (416)-978-5325
Toronto, ON  M5S 3H6, Canada             finger for PGP public key

[BProc] Reply to : mpif77 and mpif90 in Clustermatic 5 (Tomasz Perlik

[<ro...@or...>]

I have the same problem. I use Clustermatic 5 for our cluster and some pr=
ograms
( MPQC ) requires a fortran compiler. Clustermatic5 does not have any so =
I took
another mpich that had all the compilers present (mpich-p4, I don't remem=
ber
from where, but it should not be hard to find via google). When I e-maile=
d the
staff at clustermatic (www.clustermatic.org) they said that this could be=
 a
possible approach, however one has to change the libmpich.a to the one su=
pplied
with the clustermatic5. But if this does not work one has to e-mail them,
becuase one then has to try the second approach, which involves recompili=
ng the
whole thing from the begining. This has to be done to get the compilers w=
orking
for bproc type systems.
I hope this can be of some help to you.

/Sincerely
Robert Eklund, Ph D


Quoting bpr...@li...:

> Send BProc-users mailing list submissions to
> 	bpr...@li...
>
> To subscribe or unsubscribe via the World Wide Web, visit
> 	https://lists.sourceforge.net/lists/listinfo/bproc-users
> or, via email, send a message with subject or body 'help' to
> 	bpr...@li...
>
> You can reach the person managing the list at
> 	bpr...@li...
>
> When replying, please edit your Subject line so it is more specific
> than "Re: Contents of BProc-users digest..."
>
>
> Today's Topics:
>
>    1. mpif77 and mpif90 in Clustermatic 5 (Tomasz Perlik)
>
> --__--__--
>
> Message: 1
> Date: Tue, 30 Aug 2005 02:57:36 -0400
> From: Tomasz Perlik <bp...@o2...>
> To: bpr...@li...
> Subject: [BProc] mpif77 and mpif90 in Clustermatic 5
>
> Helo,
>
> I try compile my program in fortran on Clustermatic 5 and I don't have=3D=
20
> compilers such mpif77 and mpif90? I remember that in Clustermatic 4 I=3D=
20
> used mpif77. Why? Maybe I do something wrong or there is another way?
>
> Any suggestion is appreciated.
>
> --=3D20
>
> Tomasz Perlik
>
> V FDK
> Faculty of Physics, Astronomy and Informatics
> Nicolaus Copernicus University, Toru=3DF1, POLAND
>
>
>
>
>
> --__--__--
>
> _______________________________________________
> BProc-users mailing list
> BPr...@li...
> https://lists.sourceforge.net/lists/listinfo/bproc-users
>
>
> End of BProc-users Digest
>


--

[BProc] mpif77 and mpif90 in Clustermatic 5

[Tomasz P. <bp...@o2...>]

Helo,

I try compile my program in fortran on Clustermatic 5 and I don't have=20
compilers such mpif77 and mpif90? I remember that in Clustermatic 4 I=20
used mpif77. Why? Maybe I do something wrong or there is another way?

Any suggestion is appreciated.

--=20

Tomasz Perlik

V FDK
Faculty of Physics, Astronomy and Informatics
Nicolaus Copernicus University, Toru=F1, POLAND

Re: [BULK] [BProc] Linux bios loading modules on compute nodes

[Jeff R. <jra...@ln...>]

Adding:
kmod mtdcore mtdchar chipreg gen_probe jedec_probe cfi_cmdset_0002
amd76xrom

in place of your kernel modules should fix this .

I currently have this working on the hdama (I believe this is the board
you are using).

Jeff

On Thu, 2005-08-25 at 22:27 -0500, Rene Salmon wrote:
> Hi list,
> 
> We have some compute nodes with linux bios on them and we are trying to 
> get a Bproc kernel with the MTD support to boot on these nodes.  The 
> idea is that once we get this working we can use the lbflash, cmos_util 
> tools to change or update the setting on the BIOS.
> 
> I think I am missing something here.  The head node boots fine and I can 
> load the MTD modules just fine.
> 
> lsmod:
> ------
> amd76xrom               5632  0
> chipreg                 4992  3 jedec_probe,cfi_probe,amd76xrom
> map_funcs               3328  1 amd76xrom
> mtdcore                 9616  4 mtdpart,amd76xrom,mtdchar
> 
> 
> So what I did was add this to the /etc/clustermatic/node_up.conf
> here are the last few lines of that:
> -----------------------------------------
> # Put the file system together
> plugin setupfs # File system - requires kmod....
> plugin miscfiles /dev/null /dev/zero /dev/ptmx /dev/mem
> plugin miscfiles /etc/localtime /etc/ld.so.cache /tmp # copy files
> plugin miscfiles /etc/clustermatic/nsswitch.conf>/etc/nsswitch.conf
> 
> plugin vmadlib # Setup shared libraries.
> plugin nodeinfo # Make note of information about this node
> 
> #this is for nfs
> kmod nfs
> kmod lockd
> kmod sunrpc
> #this is for LINUXBIOS
> kmod mtdchar
> kmod amd76xrom
> 
> 
> When I boot the nodes up I get this boot message:
> -----------------------------------------------------------
> 
> boot: RARP: eth0 00:50:45:5C:A0:5C -> 10.0.0.4/255.0.0.0
> boot: RARP: bproc=2223; file=4711; file=/var/clustermatic/boot.img
> boot: Server IP address: 10.0.0.2
> boot: My IP address : 10.0.0.4
> boot: starting bpslave: bpslave -devi 10.0.0.2 2223
> bpslave: IO daemon started; pid=209
> bpslave: Starting new slave 0
> bpslave-0: Connecting to 10.0.0.2:2223...
> bpslave-0: Connection to 10.0.0.2:2223 up and running
> bpslave-0: Setting node number to 1
> bpslave: Master sets:
> bpslave: 0: 10.0.0.2 2223
> bcm5700: eth1 NIC Link is DOWN
> nodeup : ****amd76xrom amd76xrom_init_one(): Unable to register resource 
> 0xff
> b00000-0xffffffff - kernel bug?
> ** This is node amd76xrom window : 500000 at ffb00000
> 1 ******
> amd76xrom amd76xrom_init_one(): Unable to register resource 
> 0xffc00000-0xfffffff
> f - kernel bug?
> amd76xrom window : 400000 at ffc00000
> amd76xrom amd76xrom_init_one(): Unable to register resource 
> 0xffff0000-0xfffffff
> f - kernel bug?
> amd76xrom window : 10000 at ffff0000
> nodeup : Node setup completed successfully.
> 
> 
> Any ideas as to what I am missing? does anyone have a node_up.conf file 
> that works that I can use as a reference?
> 
> Thanks
> Rene
> 
> 
> 
> 
> 
> -------------------------------------------------------
> SF.Net email is Sponsored by the Better Software Conference & EXPO
> September 19-22, 2005 * San Francisco, CA * Development Lifecycle Practices
> Agile & Plan-Driven Development * Managing Projects & Teams * Testing & QA
> Security * Process Improvement & Measurement * http://www.sqe.com/bsce5sf
> _______________________________________________
> BProc-users mailing list
> BPr...@li...
> https://lists.sourceforge.net/lists/listinfo/bproc-users
> 

Re: [BProc] Linux bios loading modules on compute nodes

[Joshua A. <lu...@ln...>]

> Any ideas as to what I am missing? does anyone have a node_up.conf file 
> that works that I can use as a reference?

I am by no means an MTD expert, but in the past I have had MTD drivers
be very picky about the order that they are loaded in.  It looks like
you arn't calling out all the modules in nodeup.conf.  I would start by
calling out all the mtd drivers in the specific order you want them
loading in.  As for your specific errors I have no clue.

FWIW you don't have to have MTD working in order to use cmos_util, just
to actually reflash the bios.  What Mobo are you using?  I may have a
node_up.conf laying around for it.

Josh

[BProc] Linux bios loading modules on compute nodes

[Rene S. <rs...@tu...>]

Hi list,

We have some compute nodes with linux bios on them and we are trying to 
get a Bproc kernel with the MTD support to boot on these nodes.  The 
idea is that once we get this working we can use the lbflash, cmos_util 
tools to change or update the setting on the BIOS.

I think I am missing something here.  The head node boots fine and I can 
load the MTD modules just fine.

lsmod:
------
amd76xrom               5632  0
chipreg                 4992  3 jedec_probe,cfi_probe,amd76xrom
map_funcs               3328  1 amd76xrom
mtdcore                 9616  4 mtdpart,amd76xrom,mtdchar


So what I did was add this to the /etc/clustermatic/node_up.conf
here are the last few lines of that:
-----------------------------------------
# Put the file system together
plugin setupfs # File system - requires kmod....
plugin miscfiles /dev/null /dev/zero /dev/ptmx /dev/mem
plugin miscfiles /etc/localtime /etc/ld.so.cache /tmp # copy files
plugin miscfiles /etc/clustermatic/nsswitch.conf>/etc/nsswitch.conf

plugin vmadlib # Setup shared libraries.
plugin nodeinfo # Make note of information about this node

#this is for nfs
kmod nfs
kmod lockd
kmod sunrpc
#this is for LINUXBIOS
kmod mtdchar
kmod amd76xrom


When I boot the nodes up I get this boot message:
-----------------------------------------------------------

boot: RARP: eth0 00:50:45:5C:A0:5C -> 10.0.0.4/255.0.0.0
boot: RARP: bproc=2223; file=4711; file=/var/clustermatic/boot.img
boot: Server IP address: 10.0.0.2
boot: My IP address : 10.0.0.4
boot: starting bpslave: bpslave -devi 10.0.0.2 2223
bpslave: IO daemon started; pid=209
bpslave: Starting new slave 0
bpslave-0: Connecting to 10.0.0.2:2223...
bpslave-0: Connection to 10.0.0.2:2223 up and running
bpslave-0: Setting node number to 1
bpslave: Master sets:
bpslave: 0: 10.0.0.2 2223
bcm5700: eth1 NIC Link is DOWN
nodeup : ****amd76xrom amd76xrom_init_one(): Unable to register resource 
0xff
b00000-0xffffffff - kernel bug?
** This is node amd76xrom window : 500000 at ffb00000
1 ******
amd76xrom amd76xrom_init_one(): Unable to register resource 
0xffc00000-0xfffffff
f - kernel bug?
amd76xrom window : 400000 at ffc00000
amd76xrom amd76xrom_init_one(): Unable to register resource 
0xffff0000-0xfffffff
f - kernel bug?
amd76xrom window : 10000 at ffff0000
nodeup : Node setup completed successfully.


Any ideas as to what I am missing? does anyone have a node_up.conf file 
that works that I can use as a reference?

Thanks
Rene

[BProc] 2.6.11 development

[Motu <mo...@gm...>]

Hello all,
Does anyone have any information on whether the cvs builds of bproc will be=
=20
usable with the 2.6.11 kernel?

Prabhas

[BProc] bpsh environment lost

[Marcel A. <m_a...@ho...>]

I'm using a Beowulf cluster, but found that many scripts doesn't work, 
but do if directly executed. Even this simple command fails:

$ bpsh 1 /bin/bash -c 'env'
/bin/bash: env: command not found

I want to avoid including the full path to /every/ executable in the 
script (as in bpsh 1 /bin/bash -c '/usr/bin/env'). Any suggestion?

Marcel

[BProc] Re: Clustermatic's kernel-2.6.9-cm46: NMI lockup on CPU1, CPU3

[Motu <mo...@gm...>]

Hello,
I am working with a machine with 2 250 gb SATA harddrives, 2 dual-core=20
opteron 2.2 processors, 2 broadcom netxtreme 5704 gigabit ethernet network=
=20
cards, and a tyan s2881 motherboard.
When I try to boot with the kernel version 2.6.9-cm46 (the one provided wit=
h=20
clustermatic 5)--as opposed to the 2.6.11-FC4smp that came with the Fedora=
=20
Core 4 system, I get NMI lockup messages on both the CPUs.
The following is what the tail-end of hte message looks like:

> Code: 80 3f 00 7e f9 e9 8b fd ff ff e8 1e bf ec ff e9 a1 fd ff ff

 Console shuts up...
>=20
 NMI Watchdog detected LOCKUP on CPU1, registers:
> CPU1
> Modules Linked in:
> Pid:0, comm:swapper No tainted 2.6.9-cm46
> RIP:0010:[<ffffffff8032451b7>] NMI Watchdog detected LOCKUP on CPU3,=20
> registers:
> CPU3
> Modules Linked in:
> Pid:0, comm:swapper No tainted 2.6.9-cm46
> RIP:0010:[<ffffffff8032451b7>]<fffffffff80136a21>{schedulertick+385}=20


Does anyone know the cause of this problem? Anyway to deduce it?
The web seems to be useless: everyone with a NMI lockup fixed the problem=
=20
with a newer kernel, but in my case, I need to use clustermatic's kernel. I=
s=20
there any way out?

Prabhas

[BProc] RE: Broc front end host locked up

[Sun Y. <sy...@pl...>]

The lock-up only occurs if  SIGXCPU signal is sent to
a process group, killing the processes in the group
respectively with SIGXCPU seems fine.

Regards,

Yi Sun




-----Original Message-----
From: Sun Yi=20
Sent: Tuesday, August 02, 2005 5:13 PM
To: 'bpr...@li...'
Subject: Broc front end host locked up


I just ran a test sending SIGXCPU to a process group, and expected
BProc would forward the signal to the processes on remote node, however
after issuing "kill -SIGXCPU -process_group_id", the master node
got locked up, and a reboot had to be performed. I'm not sure
if this is known issue, any input will be appreciated.

Regards,

Yi Sun

Re: [BProc] python, shell, and bproc

[Peter L. <lar...@um...>]

On Aug 2, 2005, at 7:04 PM, adrian wrote:

> On 8/1/05, Peter Larkowski <lar...@um...> wrote:
>> Hello:
>>
>> I'm setting up a smallish (24 node, 48 processor) cluster to run a
>> software package called dacapo (atomic simulation package).  This
>> package consists of a binary (pile of fortran compiled against mpi) 
>> that
>> calculates the wavefunctions and some python modules to control what 
>> the
>> binaries calculate.  I have the cluster setup with clustermatic 5 and
>> the binary executes fine with mpirun.  What happens after the
>> wavefunctions are calculated (interitive solution) is python is 
>> supposed
>> to get a signal and do the next thing in the script (geometry step or
>> finish or whatever the script says to do next, etc...) but this never
>> happens and it just runs forever or until bjs effectively "kills" the 
>> job.
>>
>> If I kill the job myself, the python script on the head node just runs
>> forever (it doesn't ever seem to figure out that dacapo has died).
>>
>> I've played around with various ways of running this software 
>> (bpsh'ing
>> the python script to a node first, and then having dacapo execute - 
>> this
>> doesn't work at all which is I guess expected) and I've played messed
>> with /proc/sys/bproc/shell_hack, etc....  Just running the python 
>> script
>> on the head node and having it call a shell script which executes 
>> mpirun
>> is the closest I get, but it does what I describe above.
>>
>> I realize the people on this list probably don't have experience with
>> this software, and I'm not even sure I understand the internals
>> completely anyway (I've dug into some of the python and I'll admit 
>> some
>> of it starts to border on magic to me.....), but the basic question 
>> here
>> is does this sound like the kind of thing that bproc just doesn't 
>> handle
>> well yet?  I'm starting to get the feeling I should cut my losses and
>> just setup independent nodes with ssh keys shared, lam-mpi or mpich, 
>> and
>> either openpbs or pbs-pro.  We have access to a cluster set up this 
>> way,
>> and the software does work fine, but this clustermatic setup is nifty
>> from an admin standpoint so I'd love to make it work, but we need this
>> cluster running, so I'm starting to think I should abort and go the
>> other way now.  What do you guys think?  Sorry for the long message, 
>> but
>> I thought I should describe my problems as thoroughly as I can.
>>
>>
> Peter,
>         I would be willing to take a look at the python code for you,
> and hopefully shed a little light on what the python code is doing.
> Have you tried using the python debugger ? From you explaination, it
> sounds like perfect clustermatic material to me.
>
I've done some more testing, and I think I can get around my orginal 
problem at least to some extent, but a larger problem has presented 
itself.  Our code is run from python scripts, many of which read in 
very large input files (> 1GB).  On a traditional cluster, the jobs get 
run on the 1st node on the list of nodes assigned to that job.  On our 
bproc cluster, all the python is getting executed on the head node and 
a couple of jobs running simultaneously brings the head node to its 
knees.  Is there a way to execute python scripts on the slave nodes, 
and then have them execute the compute binaries via mpirun?  I've read 
about some bproc python bindings, but they seem very old and don't 
compile against 4.0.0pre8.  Is there a way to start the job on the head 
node, and then migrate it to the slave node or something along those 
lines?

Thanks,
Peter

RE: [BProc] syslog message

[Sun Y. <sy...@pl...>]

Eric,

Thanks for asking.=20

It is BProc 4, Linux kernel 2.6 on AMD64.

Regards,

Yi Sun


-----Original Message-----
From: Erik Hendriks [mailto:eah...@gm...]
Sent: Tuesday, August 02, 2005 5:46 PM
To: Sun Yi
Cc: BPr...@li...
Subject: Re: [BProc] syslog message


What version of Linux and BProc are you using?

- Erik

On 7/27/05, Sun Yi <sy...@pl...> wrote:
> Linux syslog contains a lot messages as shown below, any idea?
> kernel: bproc: ghost: signal: signr=3D=3D 0
>=20
> When they are presented, the linux system may hang,
> and changing ownership of a node may fail.
>=20
> Thanks in advance.
>=20
> Yi Sun
>=20
>=20
>=20
>=20
> -------------------------------------------------------
> SF.Net email is Sponsored by the Better Software Conference & EXPO =
September
> 19-22, 2005 * San Francisco, CA * Development Lifecycle Practices
> Agile & Plan-Driven Development * Managing Projects & Teams * Testing =
& QA
> Security * Process Improvement & Measurement * =
http://www.sqe.com/bsce5sf
> _______________________________________________
> BProc-users mailing list
> BPr...@li...
> https://lists.sourceforge.net/lists/listinfo/bproc-users
>

Re: [BProc] syslog message

[Erik H. <eah...@gm...>]

What version of Linux and BProc are you using?

- Erik

On 7/27/05, Sun Yi <sy...@pl...> wrote:
> Linux syslog contains a lot messages as shown below, any idea?
> kernel: bproc: ghost: signal: signr=3D=3D 0
>=20
> When they are presented, the linux system may hang,
> and changing ownership of a node may fail.
>=20
> Thanks in advance.
>=20
> Yi Sun
>=20
>=20
>=20
>=20
> -------------------------------------------------------
> SF.Net email is Sponsored by the Better Software Conference & EXPO Septem=
ber
> 19-22, 2005 * San Francisco, CA * Development Lifecycle Practices
> Agile & Plan-Driven Development * Managing Projects & Teams * Testing & Q=
A
> Security * Process Improvement & Measurement * http://www.sqe.com/bsce5sf
> _______________________________________________
> BProc-users mailing list
> BPr...@li...
> https://lists.sourceforge.net/lists/listinfo/bproc-users
>

[BProc] Broc front end host locked up

[Sun Y. <sy...@pl...>]

I just ran a test sending SIGXCPU to a process group, and expected
BProc would forward the signal to the processes on remote node, however
after issuing "kill -SIGXCPU -process_group_id", the master node
got locked up, and a reboot had to be performed. I'm not sure
if this is known issue, any input will be appreciated.

Regards,

Yi Sun

[BProc] python, shell, and bproc

[Peter L. <lar...@um...>]

Hello:

I'm setting up a smallish (24 node, 48 processor) cluster to run a 
software package called dacapo (atomic simulation package).  This 
package consists of a binary (pile of fortran compiled against mpi) that 
calculates the wavefunctions and some python modules to control what the 
binaries calculate.  I have the cluster setup with clustermatic 5 and 
the binary executes fine with mpirun.  What happens after the 
wavefunctions are calculated (interitive solution) is python is supposed 
to get a signal and do the next thing in the script (geometry step or 
finish or whatever the script says to do next, etc...) but this never 
happens and it just runs forever or until bjs effectively "kills" the job.

If I kill the job myself, the python script on the head node just runs 
forever (it doesn't ever seem to figure out that dacapo has died).

I've played around with various ways of running this software (bpsh'ing 
the python script to a node first, and then having dacapo execute - this 
doesn't work at all which is I guess expected) and I've played messed 
with /proc/sys/bproc/shell_hack, etc....  Just running the python script 
on the head node and having it call a shell script which executes mpirun 
is the closest I get, but it does what I describe above.

I realize the people on this list probably don't have experience with 
this software, and I'm not even sure I understand the internals 
completely anyway (I've dug into some of the python and I'll admit some 
of it starts to border on magic to me.....), but the basic question here 
is does this sound like the kind of thing that bproc just doesn't handle 
well yet?  I'm starting to get the feeling I should cut my losses and 
just setup independent nodes with ssh keys shared, lam-mpi or mpich, and 
either openpbs or pbs-pro.  We have access to a cluster set up this way, 
and the software does work fine, but this clustermatic setup is nifty 
from an admin standpoint so I'd love to make it work, but we need this 
cluster running, so I'm starting to think I should abort and go the 
other way now.  What do you guys think?  Sorry for the long message, but 
I thought I should describe my problems as thoroughly as I can.

Thanks for your input.

-Peter

[BProc] syslog message

[Sun Y. <sy...@pl...>]

Linux syslog contains a lot messages as shown below, any idea?
kernel: bproc: ghost: signal: signr=3D=3D 0

When they are presented, the linux system may hang,
and changing ownership of a node may fail.

Thanks in advance.

Yi Sun

Re: [BProc] bproc crash please help!!

[Rene S. <rs...@tu...>]

Hi List,

The problem is somewhat reprodusable.  The cluster has crashed a couple 
of times in the past few days with similar messages.  Following is the 
syslog entry right before the crash for today.  The cluster consists of 
40 dual opteron nodes.
Here is what we do to reproduce the crash:

1)The cluster boots fine and all the nodes come up no problem.
2)We then procced to queue up some mpi and non mpi jobs to run.
3)After a few hours of running the jobs the cluster becomes unresponsibe 
meaning we can't get a prompt at the console, we can't ssh in we can't 
get anywhere or do anything other than power cycle the whole thing.

Is there any way to get more debug info out of bproc so that I can get 
a clue as to where to start looking at what might be causing this problem?

thank you for any advice/help on this.

Rene


> Jul 26 18:14:58 aresnfs-frontend.ccsnfs.edu kernel: bproc: ghost: signal: signr 
> == 0 
> Jul 26 18:14:58 aresnfs-frontend.ccsnfs.edu last message repeated 577 times 
> Jul 26 18:14:58 aresnfs-frontend.ccsnfs.edu kernel: proc: ghproc: ghost: siproc:
>  ghosproc: ghost: sigproc: gproc: ghost: sigproc: ghost: sigproc: ghost: sigproc
> : ghost: sigproc: ghost: siproc: ghost: sigproc: ghostproc: ghost: sigproc: ghos
> t: sigproc: ghost: sigproc: ghost: sigproc: ghost: sigproc: ghost: sigproc: ghos
> tproc: ghost: sigproc: ghost: signal: signr == 0 
> Jul 26 18:14:58 aresnfs-frontend.ccsnfs.edu kernel: bproc: ghost: signal: signr 
> == 0 
> Jul 26 18:14:59 aresnfs-frontend.ccsnfs.edu last message repeated 3631 times 
> Jul 26 18:14:59 aresnfs-frontend.ccsnfs.edu kernel: proc: ghost: signal: signr =
> = 0 
> Jul 26 18:14:59 aresnfs-frontend.ccsnfs.edu kernel: bproc: ghost: signal: signr 
> == 0 
> Jul 26 18:14:59 aresnfs-frontend.ccsnfs.edu last message repeated 3639 times 
> Jul 26 18:14:59 aresnfs-frontend.ccsnfs.edu kernel: proc: ghost: signal: signr =
> = 0 
> Jul 26 18:14:59 aresnfs-frontend.ccsnfs.edu kernel: bproc: ghost: signal: signr 
> == 0 




Erik Hendriks wrote:
> I've never seen a problem like that.  Is it reproducable?
> 
> - Erik
>