bproc-users — General discussion about BProc.

[BProc] bproc-4.0.0pre6 patch error on kernel 2.6.7.

[Gustavo G. M. <gu...@ma...>]

Hi everybody again,

Luke has helped me and this message is the continuation of the problems.

Luke, as you said, I installed the kernel 2.6.7 and recompiled it. Until this,
Ok. The kernel 2.6.7 was compiled.

The next step was apply the bproc-4.0.0pre6.

I execute the following command:

patch -f -p1 < ../bproc-4.0.0pre6/patches/bproc-4.0.0pre6

But, again, I got some errors. Even this, I tried to recompile de kernel, but
without success.

A question: Can I edit the files where the patch failed, looking for the block
code in the patch bproc-4.0.0pre6 and pasting it in the kernel file?

The errors applying the patch are:

patching file arch/i386/kernel/i386_ksyms.c
Hunk #1 FAILED at 206.
1 out of 1 hunk FAILED -- saving rejects to file
arch/i386/kernel/i386_ksyms.c.rej

patching file arch/i386/kernel/process.c
Hunk #1 FAILED at 36.
1 out of 3 hunks FAILED -- saving rejects to file arch/i386/kernel/process.c.rej

patching file arch/i386/kernel/traps.c
Hunk #1 FAILED at 61.
1 out of 1 hunk FAILED -- saving rejects to file arch/i386/kernel/traps.c.rej

patching file arch/x86_64/kernel/x8664_ksyms.c
Hunk #1 FAILED at 219.
1 out of 1 hunk FAILED -- saving rejects to file
arch/x86_64/kernel/x8664_ksyms.c.rej

patching file kernel/sched.c
Hunk #1 FAILED at 40.
1 out of 6 hunks FAILED -- saving rejects to file kernel/sched.c.rej

--
Atenciosamente,
Gustavo Gobi Martinelli
Linux User# 270627

[BProc] Workshop: Clustermatic - An Innovative Approach To Cluster Computing

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

	From: 	  gw...@la...
	Subject: 	Workshop - Clustermatic: An Innovative Approach To Cluster 
Computing
	Date: 	September 28, 2004 6:12:31 PM MDT
	To: 	  cc...@la..., cc...@la...

Clustermatic: An Innovative Approach To Cluster Computing

The 2004 LACSI Symposium will present this workshop on October 12.  
This workshop will give participants extensive, hands-on experience, 
installing, configuring and using the Clustermatic software suite on a 
real cluster computer system.

One can register for the workshop only, or the entire Symposium, at 
http://lacsi.lanl.gov/symposium, then click on "Registration".  After 
October 5, registration is available onsite at the Eldorado Hotel in 
Santa Fe.

AGENDA

8:00-9:00    Registration and Continental Breakfast

9:00-9:30    Introduction to Clustermatic, Greg Watson, Cluster 
Research Team, LANL

9:30-10:30  BProc & Beoboot, Erik Hendriks, Cluster Research Team, LANL

10:30-11:00  Break

11:00-11:30  BProc & Beoboot continued..., Erik Hendriks, Cluster 
Research Team, LANL

11:30-12:30  LinuxBIOS, Ron Minnich, Cluster Research Team Lead, LANL

12:30-2:00   Lunch (on your own)

2:00-3:00    Filesystems, Ron Minnich, Cluster Research Team Lead, LANL

3:00-3:30    Supermon, Matt Sottile, Cluster Research Team, LANL

3:30-4:00    Break

4:00-4:30    BProc Job Scheduler (BJS), Matt Sottile, Cluster Research 
Team, LANL

4:30-5:30    Using MPI, Matt Sottile, Cluster Research Team, LANL

[BProc] tg3: no version for kfree found

[Ted S. <tsa...@cr...>]

Hi,

I work to install clustermatic4 on an opteron cluster running SuSe 
Server9 for AMD64 (Tian B2882 Transport GX28). I compiled kernel 2.6.7 
with modules from  http://sf.net/projects/bproc (thanks Erik). At boot 
phase 2 I get an error:

boot: Install module e100
e100: Intel PRO/100 driver 3.0.18
socket: Address family not supported by protocol

If if disable the hardware (I don't use it) I get similar error for the 
gigabit interface:

tg3: no version for kfree found: kernel tainted
socket: Address family not supported by protocol

What do I do about this kind of errors?
I'll highly appreciate any help.
Thanks,
-- 
Ted

RE: [BProc] BPROC error on Kernel 2.6.5

[Gustavo G. M. <gu...@ma...>]

Luke,

Thanks for your answer.

I have a doubt: the patch bproc-4.0.0pre6 is not only for the kernel 2.6.7?

Because my kernel is 2.6.5.

--
Atenciosamente,
Gustavo Gobi Martinelli
Linux User# 270627


Citando Luke Palmer <lop...@wi...>:

> Gustavo,
>
> The build procedure in 2.6 is probably different than what you're used to.
> I'm not sure if this would help your problem or not.  I can tell you that
> "make bzImage" is depreciated in favor of just "make", but reading the docs
> would probably be a good idea.  Try it that way with a clean source tree,
> and let us know what happens.
>
> Also, why are you using pre4?  pre6 is available, and will very likely be
> cleaner.  pre6 on FC2 compiles fine for me.
>
> -Luke
>
>
> -----Original Message-----
> From: bpr...@li...
> [mailto:bpr...@li...] On Behalf Of Gustavo Gobi
> Martinelli
> Sent: Tuesday, September 28, 2004 8:07 AM
> To: bpr...@li...
> Cc: clu...@gr...
> Subject: [BProc] BPROC error on Kernel 2.6.5
>
> Hi everybody,
>
> I´m trying to install a beowulf cluster and I applied the bproc-4.0.0pre4
> patch
> on my kernel 2.6.5 (Fedora Core 2).
>
> But when I try to recompile my kernel, I find an error exactly as below.
> After
> the "make bzImage" command:
>
> "arch/i386/kernel/ptrace.c: At top level:
> arch/i386/kernel/ptrace.c:529: warning: type defaults to `int' in
> declaration of
> `EXPORT_SYMBOL'
> arch/i386/kernel/ptrace.c:529: warning: parameter names (without types) in
> function declaration
> arch/i386/kernel/ptrace.c:529: warning: data definition has no type or
> storage
> class
> make[1]: ** [arch/i386/kernel/ptrace.o] Erro 1
> make: ** [arch/i386/kernel] Erro 2"
>
> What do I have to do about this error? Can I edit the file ptrace.c to
> verify
> this errors?
>
> Another thing to know: I only execute the "make bzImage" command, because
> the
> recompilation for kernel 2.6.5 change some old command like "make dep". Is
> there ok?
>
> --
> Atenciosamente,
> Gustavo Gobi Martinelli
> Linux User# 270627
>
>
>
>
>
>
> -------------------------------------------------------
> This SF.Net email is sponsored by: YOU BE THE JUDGE. Be one of 170
> Project Admins to receive an Apple iPod Mini FREE for your judgement on
> who ports your project to Linux PPC the best. Sponsored by IBM.
> Deadline: Sept. 24. Go here: http://sf.net/ppc_contest.php
> _______________________________________________
> BProc-users mailing list
> BPr...@li...
> https://lists.sourceforge.net/lists/listinfo/bproc-users
>
>

RE: [BProc] BPROC error on Kernel 2.6.5

[Luke P. <lop...@wi...>]

Gustavo,

The build procedure in 2.6 is probably different than what you're used =
to.
I'm not sure if this would help your problem or not.  I can tell you =
that
"make bzImage" is depreciated in favor of just "make", but reading the =
docs
would probably be a good idea.  Try it that way with a clean source =
tree,
and let us know what happens.

Also, why are you using pre4?  pre6 is available, and will very likely =
be
cleaner.  pre6 on FC2 compiles fine for me.

-Luke


-----Original Message-----
From: bpr...@li...
[mailto:bpr...@li...] On Behalf Of Gustavo =
Gobi
Martinelli
Sent: Tuesday, September 28, 2004 8:07 AM
To: bpr...@li...
Cc: clu...@gr...
Subject: [BProc] BPROC error on Kernel 2.6.5

Hi everybody,

I=B4m trying to install a beowulf cluster and I applied the =
bproc-4.0.0pre4
patch
on my kernel 2.6.5 (Fedora Core 2).

But when I try to recompile my kernel, I find an error exactly as below.
After
the "make bzImage" command:

"arch/i386/kernel/ptrace.c: At top level:
arch/i386/kernel/ptrace.c:529: warning: type defaults to `int' in
declaration of
`EXPORT_SYMBOL'
arch/i386/kernel/ptrace.c:529: warning: parameter names (without types) =
in
function declaration
arch/i386/kernel/ptrace.c:529: warning: data definition has no type or
storage
class
make[1]: ** [arch/i386/kernel/ptrace.o] Erro 1
make: ** [arch/i386/kernel] Erro 2"

What do I have to do about this error? Can I edit the file ptrace.c to
verify
this errors?

Another thing to know: I only execute the "make bzImage" command, =
because
the
recompilation for kernel 2.6.5 change some old command like "make dep". =
Is
there ok?

--
Atenciosamente,
Gustavo Gobi Martinelli
Linux User# 270627






-------------------------------------------------------
This SF.Net email is sponsored by: YOU BE THE JUDGE. Be one of 170
Project Admins to receive an Apple iPod Mini FREE for your judgement on
who ports your project to Linux PPC the best. Sponsored by IBM.
Deadline: Sept. 24. Go here: http://sf.net/ppc_contest.php
_______________________________________________
BProc-users mailing list
BPr...@li...
https://lists.sourceforge.net/lists/listinfo/bproc-users

[BProc] BPROC error on Kernel 2.6.5

[Gustavo G. M. <gu...@ma...>]

Hi everybody,

I´m trying to install a beowulf cluster and I applied the bproc-4.0.0pre4 patch
on my kernel 2.6.5 (Fedora Core 2).

But when I try to recompile my kernel, I find an error exactly as below. After
the "make bzImage" command:

"arch/i386/kernel/ptrace.c: At top level:
arch/i386/kernel/ptrace.c:529: warning: type defaults to `int' in declaration of
`EXPORT_SYMBOL'
arch/i386/kernel/ptrace.c:529: warning: parameter names (without types) in
function declaration
arch/i386/kernel/ptrace.c:529: warning: data definition has no type or storage
class
make[1]: ** [arch/i386/kernel/ptrace.o] Erro 1
make: ** [arch/i386/kernel] Erro 2"

What do I have to do about this error? Can I edit the file ptrace.c to verify
this errors?

Another thing to know: I only execute the "make bzImage" command, because the
recompilation for kernel 2.6.5 change some old command like "make dep". Is
there ok?

--
Atenciosamente,
Gustavo Gobi Martinelli
Linux User# 270627

[BProc] RE: bproc patch for 2.6.4-52 in Suse 9.1 pro amd64

[YhLu <Yh...@ty...>]

It does not with kernel 2.6.8.1 in kernel.org too.

YH

-----Original Message-----
From: er...@he... [mailto:er...@he...] 
Sent: Thursday, September 23, 2004 7:03 AM
To: YhLu
Cc: bpr...@li...
Subject: Re: bproc patch for 2.6.4-52 in Suse 9.1 pro amd64

On Wed, Sep 22, 2004 at 04:07:44PM -0700, YhLu wrote:
> Erik,
>  
> I used the patch in bpro-4.0.0pre6 to the Kernel 2.6.4-52 in Suse 9.1 for
> AMD64.
> Got some rej.

The patches I generate are against the stock kernels from kernel.org.
Vendors like SuSE and Red Hat apply literally hundreds of patches to
their kernels.  My patches never apply cleanly after that.  I
recommend just using a kernel.org kernel.  Otherwise it's up to you to
resolve the conflicts.

- Erik

[BProc] Re: bproc patch for 2.6.4-52 in Suse 9.1 pro amd64

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

On Wed, Sep 22, 2004 at 04:07:44PM -0700, YhLu wrote:
> Erik,
>  
> I used the patch in bpro-4.0.0pre6 to the Kernel 2.6.4-52 in Suse 9.1 for
> AMD64.
> Got some rej.

The patches I generate are against the stock kernels from kernel.org.
Vendors like SuSE and Red Hat apply literally hundreds of patches to
their kernels.  My patches never apply cleanly after that.  I
recommend just using a kernel.org kernel.  Otherwise it's up to you to
resolve the conflicts.

- Erik

Re: [BProc] depmod: QM_MODULES: Function not implemented

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

On Mon, Sep 20, 2004 at 08:22:46AM -0400, Ted Sariyski wrote:
> Hi,
> I try to install clustermatic4 on a opteron cluster running SuSe Server9
> for AMD64 (Tian B2882 Transport GX28). While installing rpms I get
> errors for bproc-modules and v9fs-mounter-smp:
> 
> #> depmod: QM_MODULES: Function not implemented
> 
> Does  anybody else had  this problem?

The Clustermatic 4 RPMS are for Linux 2.4.  A lot has changed between
Linux 2.4 and Linux 2.6.  In order to use 2.6, I'll have to build from
newer sources.  (available at http://sf.net/projects/bproc)

- Erik

Re: [BProc] depmod: QM_MODULES: Function not implemented

[Steven J. <py...@li...>]

Greetings,

The kernel's module API changed between 2.4 and 2.6. You'll need to
update your module utilities to use the new syscalls.

G'day,
sjames


||||| |||| |||||||||||||  |||
by Linux Labs International, Inc.
   Steven James, CTO

55 Marietta Street
Suite 1830
Atlanta, Ga 30303
866 824 9737 support


On Mon, 20 Sep 2004, Ted Sariyski wrote:

> Hi,
> I try to install clustermatic4 on a opteron cluster running SuSe Server9
> for AMD64 (Tian B2882 Transport GX28). While installing rpms I get
> errors for bproc-modules and v9fs-mounter-smp:
>
> #> depmod: QM_MODULES: Function not implemented
>
> Does  anybody else had  this problem?
>
> Thanks,
> Ted
>
>
>
> -------------------------------------------------------
> This SF.Net email is sponsored by: YOU BE THE JUDGE. Be one of 170
> Project Admins to receive an Apple iPod Mini FREE for your judgement on
> who ports your project to Linux PPC the best. Sponsored by IBM.
> Deadline: Sept. 24. Go here: http://sf.net/ppc_contest.php
> _______________________________________________
> BProc-users mailing list
> BPr...@li...
> https://lists.sourceforge.net/lists/listinfo/bproc-users
>

[BProc] depmod: QM_MODULES: Function not implemented

[Ted S. <tsa...@cr...>]

Hi,
I try to install clustermatic4 on a opteron cluster running SuSe Server9
for AMD64 (Tian B2882 Transport GX28). While installing rpms I get
errors for bproc-modules and v9fs-mounter-smp:

#> depmod: QM_MODULES: Function not implemented

Does  anybody else had  this problem?

Thanks,
Ted

Re: [BProc] bjs and ulimits

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

On Wed, Sep 15, 2004 at 07:59:46PM +0100, Miguel Dias Costa wrote:
> Hello all
> 
> does bjs set a stack ulimit to 8192? 
> 
> We verify that bpsh migrates ulimits to the nodes but with bjssub, there
> is always this stack limit.

bjs doesn't do anything with ulimits.  If you're running in batch
mode, you'll probably get whatever bjs itself gets.  In interactive
mode, your job will get whatever you've got in your shell.

bjs really should deal with the ulimits.  I was thinking that bjs
should actually get PAM in the loop when starting jobs so that things
like pam_limits can work too.

- Erik

[BProc] bjs and ulimits

[Miguel D. C. <mc...@fc...>]

Hello all

does bjs set a stack ulimit to 8192?=20

We verify that bpsh migrates ulimits to the nodes but with bjssub, there
is always this stack limit.


--=20
Miguel Dias Costa <mc...@fc...>
Centro de F=EDsica do Porto

Re: [BProc] System crash destroyed contents of node disks - why?

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

On Wed, Sep 08, 2004 at 10:43:17AM +0200, Peter Englmaier wrote:
> Hi, we had a 'Scyld' Cluster running for years until some disk
> on the master died. Some MPI program and other 'normal' programs
> were running. After the disk crash the bproc system was hanging
> and after a reboot all nodes didn't come up. Finally,
> we installed RH9+Clustermatic 4. Main reason for this: we
> had no good documentation aboud scyld and the old sysadmin had 
> left the institute. Setting up clustermatic was quite easy.
> 
> Nodes had scratch disks with three partitions: Scyld boot
> partition (actually an ext2 partition), swap, and /scratch.
> All three partitions where destroyed - I could not mount any of them.
> On about 10 of 22 nodes! Even swap was not recognized as such, although
> fdisk reported all partitions. I suspect something was going wrong
> with the kernel part of bproc. Is this possible?

BProc doesn't get its fingers in any of the kernel's file system stuff
so I think it's unlikely that BProc would induce file system problems.
I suppose it's possible if something went crazy and started scribbling
on memory.  Personally, I've never seen a problem like that with
BProc.  I can't speak for Scyld though - they've modified BProc and I
don't know the details of what they've done.

Personally, I would pull out one of the disks to see if they work on
some other stand alone system.  If they do then I'd start picking
through the node setup process to see what's going on there.

- Erik

Re: [BProc] Start processes in master with mpirun?

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

On Thu, Sep 09, 2004 at 09:41:41AM -0700, J S wrote:
> 
> We have a 3 computer cluster (each one a dual AMD), 
> and wish to use them all to do work. We have one of
> them configured as a master, and the other 2 as slaves
> (nodes 0-1). 
> 
> Executing
> 
> NODES=-1,0,1 mpirun -np 6 -s --p4 ./cpi
> Could not get number of cpus for node -1, assuming 1
> Not enough nodes to allocate all processes
> 
> (It works well with NODES=0,1 for -np 4)
> 
> We are using Bproc and MPICH w/ clustermatic. How can
> I tell mpirun to start 2 processes in the master node?

I'm afraid it's busted.  There's a -l switch but it doesn't mix well
with using the NODES string.  I guess we never fixed it because it
doesn't come up much here.  If you decide to fix it, please send a
patch.

- Erik

[BProc] Start processes in master with mpirun?

[J S <joa...@ya...>]

We have a 3 computer cluster (each one a dual AMD), 
and wish to use them all to do work. We have one of
them configured as a master, and the other 2 as slaves
(nodes 0-1). 

Executing

NODES=-1,0,1 mpirun -np 6 -s --p4 ./cpi
Could not get number of cpus for node -1, assuming 1
Not enough nodes to allocate all processes

(It works well with NODES=0,1 for -np 4)

We are using Bproc and MPICH w/ clustermatic. How can
I tell mpirun to start 2 processes in the master node?

Thanks in advance,
João Silva



		
__________________________________
Do you Yahoo!?
Yahoo! Mail - 50x more storage than other providers!
http://promotions.yahoo.com/new_mail

[BProc] System crash destroyed contents of node disks - why?

[Peter E. <Pet...@un...>]

Hi, we had a 'Scyld' Cluster running for years until some disk
on the master died. Some MPI program and other 'normal' programs
were running. After the disk crash the bproc system was hanging
and after a reboot all nodes didn't come up. Finally,
we installed RH9+Clustermatic 4. Main reason for this: we
had no good documentation aboud scyld and the old sysadmin had 
left the institute. Setting up clustermatic was quite easy.

Nodes had scratch disks with three partitions: Scyld boot
partition (actually an ext2 partition), swap, and /scratch.
All three partitions where destroyed - I could not mount any of them.
On about 10 of 22 nodes! Even swap was not recognized as such, although
fdisk reported all partitions. I suspect something was going wrong
with the kernel part of bproc. Is this possible?

Best, Peter.

-- 
Astron. Institut Uni Basel, Venusstr. 7, CH-4102 Binningen, Switzerland
Phone: +41 (61) 2055-434, fax: -455, http://www.astro.unibas.ch/~ppe/

Re: [BProc] bproc connections

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

On Mon, Sep 06, 2004 at 07:41:13PM +0200, Florian Bruckner wrote:
> can anybody tell me how bproc slave is connected with the bpmaster?
> 
> as I understood it the daemons communicate with there kernel module via 
> a filedescriptor which the get by a system call. And all communication 
> between master and slave is done via a user space socket connection. 
> Requests sent from the master are interpreted by the slave or passed on 
> to the kernel (masqfs).
> 
> But why are there sockets in kernel mode? i think they are only used for 
> moving processes around. But why aren't they also transfered via the 
> user space sockets?
> 
> Is there any documentation around that describes all connections used by 
> bproc?

The best documentation on the guts is probably here:

http://public.lanl.gov/cluster/papers/papers/hendriks-ics02.pdf

It talks a little about connections.  In a nutshell, the kernel stuff
passes messages to the daemons.  The slaves all have a tcp connection
the master.  These connections are only used to pass small messages
around.  When a process wants to move, a messages is sent to the
destination telling it that it wants to move.  The destination
establishes a new TCP connection back to the source.  The actual
process data is sent on that connection.  The kernel code establishes
this new connection.  The kernel sets up that connection for security
reasons.  That way user space processes can't poke weird stuff in
there.

- Erik

[BProc] bproc connections

[Florian B. <flo...@ao...>]

can anybody tell me how bproc slave is connected with the bpmaster?

as I understood it the daemons communicate with there kernel module via 
a filedescriptor which the get by a system call. And all communication 
between master and slave is done via a user space socket connection. 
Requests sent from the master are interpreted by the slave or passed on 
to the kernel (masqfs).

But why are there sockets in kernel mode? i think they are only used for 
moving processes around. But why aren't they also transfered via the 
user space sockets?

Is there any documentation around that describes all connections used by 
bproc?

thanks in advance
FloB
flo...@ao...

Re: [BProc] partitioning, scyld

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

On Sat, Sep 04, 2004 at 01:50:23PM -0600, Michal Jaegermann wrote:
> On 2004/Aug/20 I posted a message with a subject "Note on modules on
> nodes (NFS support in particular)" which included a sample script.
> Techniques demonstrated there apply here as well.
> 
>    Michal

Here is a sample script that I put together for a system last year and
is probably the progenator of Daryl's disk formating script.  You call
it with 'format_disk $BSH_NODERAGE'.  If you are familure with fdisk,
the gobledygook that you feed it through stdin will make sense :)

Josh

Re: [BProc] partitioning, scyld

[Michal J. <mi...@ha...>]

On Sat, Aug 28, 2004 at 03:01:02PM -0500, Luke Palmer wrote:
> I can't just do 'bpsh X fdisk' because the /dev filesystem isn't
> present on slave nodes.

There is, although various special files you would like to have
in the specific situation may be not there; but nothing prevents
you from adding these.  A simple way is to use 'plugin miscfiles'
directive in your node_up.conf and add whatever devices you would
like to see.

Alternatively something in that style will do (assuming that you
want all /dev/hda* on nodes - as an example ):

find /dev -name 'hda*' | cpio -o -c --quiet | \
  bpsh $node cpio --quiet -imd

for every "$node" you want that to happen.  This can be run from
'node_up' script, or later, or separately.  Whatever you
need/desire.  'tar' instead of 'cpio' will do as well as GNU tar
handles special device files too and there are still some other
ways to accomplish that goal.

> So- what's a good way to partition?

'sfdisk' run on nodes is handy as you can feed it a requested
partitioning scheme from stdin.  You may need some additional
libraries for that and 'ldd' will tell you what they are.  Add them
to your configuration or copy to nodes similar like the above and
you are set.  Scripting that is a good idea .:-)   For a specific
situation that will be a few lines; things are getting longer
if you are trying to be general.

You will likely need still some other shared libraries if you
would want to set some file systems, or run 'mkswap', on
freshly created disk partitions.

On 2004/Aug/20 I posted a message with a subject "Note on modules on
nodes (NFS support in particular)" which included a sample script.
Techniques demonstrated there apply here as well.

   Michal

Re: [BProc] CM4 node_up failure

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

On Tue, Aug 24, 2004 at 04:01:23PM -0700, Vipul Deokar wrote:
> Folks, 
> 
> I have a configuration of 4 identical "compute" nodes
> (with disks) and 1 slightly different "master" node.
> The master node has a slightly more powerful CPU, more
> RAM, an additonal 10/100 NIC interface to connect to
> extranet, a CDROM drive on secondary IDE, a more
> recent verison of BIOS firmware.
> 
> With Red Hat 9 (runlevel 3) and Clustermatic4 i386
> RPMS installed on computeNode#1 as master, I can build
> a cluster with the other 4 nodes as slaves. (I see one
> node rebooting every 5-6 minutes that I need to
> debug). I am using this cluster currently.
> 
> 
> However, with RedHat 9 (runlevel 3 or 5; more
> packages) and Clustermatic4 i386 RPMS installed on the
> "master", I cannot get any of the "compute" nodes up
> as slaves. The node_up script fails on all nodes with
> :
>  nodeup    : Starting 1 child processes.
>  nodeup    : Finished creating child processes.
>  nodeup    : I/O error talking to child
>  nodeup    : Child process for node 1 died with signal
>  4
> 
> 
> The same config, node_up, config.boot scripts execute
> in both cluster configuration attempts - one succeeds
> and the other fails. Any insight why this would
> happen?

Signal 4 = SIGILL.  This usually happens when the front end node is
some newer rev of processor than the back end nodes.  Red Hat is
pretty good about installing the best glibc it can on the front
end. (e.g. install the i686 version instead of the i386 one)

I usually see something like this when the slave node is some other
CPU type that doesn't qualify as i686.  The issue is really that the
library gets loaded on the front end and those instructions turn out
to not be valid on the destination node.

My work around is to downgrade the glibc on the front end to one that
will work on all nodes.  In other words, load the i386 one instead of
the i686 one.

- Erik

Re: [BProc] partitioning, scyld

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

On Sat, Aug 28, 2004 at 03:01:02PM -0500, Luke Palmer wrote:
> Hello,
> 
> This may seem a dumb/obvious question, but I'm trying to figure out how
> to easily partition disks on slave nodes.  I can't just do 'bpsh X
> fdisk' because the /dev filesystem isn't present on slave nodes.  So-
> what's a good way to partition?
> 
> Of course I read about beofdisk after some google searches.  Now, I know
> scyld has always liked to more or less hide the ways of getting their
> software for free, but I can't find scyld ANYWHERE.  I recall that
> penguin computing has something to do with it now- poked around there
> too with no luck.  Can any one comment of the fate of scyld?

I have no idea about scyld.

Daryl's disk partitioning script is newer so I figure it's more likely
to work but I do have an old mirror of their FTP site from about 1.5
yrs ago.  I've been keeping it around since Becker is in the habit of
accusing us of IP theft.  Email me directly if you want an old beoboot
source RPM or something.  The beofdisk script is in there and it's got
a nice big GPL license banner on it so it's fair game to pass around.

- Erik

Re: [BProc] child processes dying

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

On Tue, Aug 31, 2004 at 11:33:12AM -0500, Luke Palmer wrote:
> Hello,
> 
> I've noticed some daemons having trouble running on bproc nodes, usually
> complaining about child processes dying.  I'm running pre6 with 2.6.7,
> libraries via an NFS mount.
> 
> Attached is a simple pthreads test program.  It runs fine on the master,
> but segfaults when running via bpsh.  I've attached the last few lines
> of an strace of the test program.  Looks fishy.
> 
> Would others mind trying this test program, and commenting on their
> findings?

pthreads has been a sore point for a long time.  However, I think
those days might be over.  *cross fingers*

Grab the latest 'n' greatest bproc from CVS on sourceforge.net.  It's
got a lot of fixes for threading related stuff in there.  I think I'm
going to have a tagged version again pretty soon.  Also, make sure
that you're using NPTL and not the old linuxthreads stuff.  I've done
testing with NPTL only at this point.  If linuxthreads doesn't work
I'm prepared to say "use NPTL".

Note that you can't migrate a multithreaded thing at this point.  The
result of that is undefined.

- Erik

[BProc] child processes dying

[Luke P. <lop...@wi...>]

Hello,

I've noticed some daemons having trouble running on bproc nodes, usually
complaining about child processes dying.  I'm running pre6 with 2.6.7,
libraries via an NFS mount.

Attached is a simple pthreads test program.  It runs fine on the master,
but segfaults when running via bpsh.  I've attached the last few lines
of an strace of the test program.  Looks fishy.

Would others mind trying this test program, and commenting on their
findings?

Thanks
-Luke

clone(child_stack=0x40817b08,
flags=CLONE_VM|CLONE_FS|CLONE_FILES|CLONE_SIGHAND|CLONE_THREAD|CLONE_SYSVSEM|CLONE_SETTLS|CLONE_PARENT_SETTID|CLONE_CHILD_CLEARTID|CLONE_DETACHED, parent_tidptr=0x40817bf8, {entry_number:6, base_addr:0x40817bb0, limit:1048575, seg_32bit:1, contents:0, read_exec_only:0, limit_in_pages:1, seg_not_present:0, useable:1}, child_tidptr=0x40817bf8) = -1 EFAULT (Bad address)
--- SIGSEGV (Segmentation fault) @ 0 (0) ---
+++ killed by SIGSEGV +++



#include <iostream>
#include <pthread.h>
#include <stdlib.h>

using namespace std;

void *task(void *arg) {
  for (;;) { 
    cout << (char *)arg;
    cout.flush();
  }
  return NULL;
}

int main() {
  pthread_t t1;
  if ( pthread_create(&t1, NULL, task, (void *)"1") != 0 ) {
    cout << "pthread_create() error" << endl;
    abort();
  }
  task((void *)"2");
}