Re: Error with MALLOC_CHECK_ enabled

Hi, Phil,

I upgraded to 2.7.1 and no longer see the error. Thanks.

Wei-keng

On Tue, 29 Apr 2008, Phil Carns wrote:

> Hi Wei-keng,
> 
> I think this has probably been fixed in 2.7.1 if you are able to try 
> that.  I just tried to reproduce your problem using that version and 
> everything seemed to be Ok.
> 
> -Phil
> 
> Wei-keng Liao wrote:
> > I am running a simple MPI-IO program (attached) on PVFS 2.7.0.
> > When I set the environment variable MALLOC_CHECK_ to 1 or above, I got
> > error messages below: (No error if MALLOC_CHECK_ is not set and if file name
> > has "ufs:" prefix)
> > 
> > % mpicc -g pvfs2_bug.c -o pvfs2_bug
> > % setenv MALLOC_CHECK_ 3
> > % mpiexec -gdb -n 4 pvfs2_bug
> > malloc: using debugging hooks
> > malloc: using debugging hooks
> > 1: malloc: using debugging hooks
> > 2: malloc: using debugging hooks
> > 0: malloc: using debugging hooks

pvfs2-client problems in 2.7.1

We upgraded our PVFS to 2.7.1 on Monday and are having lots of
problems with pvfs2-client hanging if somebody executes a command
while sitting on PVFS and using a relative path to the executable.
Example:
  cd /pvfs/support; /pvfs/support/iozone -a
would work fine, while
  cd /pvfs/support; ./iozone -a
would hand the pvfs2-client on the current system.
A restart of the pvfs2-client fixes the problem, but in some cases
the executable is not killable afterward.

Our compute nodes are running a 2.6.15 kernel.

Thanks,
Jan

--

-- 
Jan Lindheim		 lindheim <at> cacr.caltech.edu
M.S. 158-79 Caltech
Pasadena, CA 91125 USA   
Tel: (626) 395-3926, FAX: (626) 584-5917

Re: building 2.7.1 on linux 2.6.25

Andrew Pochinsky <avp <at> MIT.EDU> writes:

> 
> Dear All,
> 	I'm wondering if anyone has managed to build kmod with the latest  
> kernel? I'm getting curious errors, e.g.,
> 
> test:pvfs-2.7.1 # make kmod
> /bin/sh: -c: line 0: syntax error near unexpected token `cat'
> /bin/sh: -c: line 0: `echo(cat /dev/null;   echo kernel//dist/ 
> pvfs2/2.6.25/pvfs-2.7.1/src/kernel/linux-2.6/pvfs2.ko;) > /dist/ 
> pvfs2/2.6.25/pvfs-2.7.1/src/kernel/linux-2.6/modules.order'
> make[3]: *** [/dist/pvfs2/2.6.25/pvfs-2.7.1/src/kernel/linux-2.6/ 
> modules.order] Error 2
> make[2]: *** [_module_/dist/pvfs2/2.6.25/pvfs-2.7.1/src/kernel/ 
> linux-2.6] Error 2
> make[1]: *** [default] Error 2
> make: *** [just_kmod] Error 2
>

I also confronted the problems exactly like yours, under the same configuration 
of linux 2.6.25 and PVFS2 2.7.1. And I have not figure it out till now. Waiting 
for the solution~

Thanks~

Re: Re: building 2.7.1 on linux 2.6.25

Hi Liu,

The fix for this is to make a small modification to the 2.6.25 kernel  
code.  Please see:

http://www.beowulf-underground.org/pipermail/pvfs2-users/2008-April/002382.html
-sam

On May 8, 2008, at 4:32 AM, Liu Yang wrote:

> Andrew Pochinsky <avp <at> MIT.EDU> writes:
>
>>
>> Dear All,
>> 	I'm wondering if anyone has managed to build kmod with the latest
>> kernel? I'm getting curious errors, e.g.,
>>
>> test:pvfs-2.7.1 # make kmod
>> /bin/sh: -c: line 0: syntax error near unexpected token `cat'
>> /bin/sh: -c: line 0: `echo(cat /dev/null;   echo kernel//dist/
>> pvfs2/2.6.25/pvfs-2.7.1/src/kernel/linux-2.6/pvfs2.ko;) > /dist/
>> pvfs2/2.6.25/pvfs-2.7.1/src/kernel/linux-2.6/modules.order'
>> make[3]: *** [/dist/pvfs2/2.6.25/pvfs-2.7.1/src/kernel/linux-2.6/
>> modules.order] Error 2
>> make[2]: *** [_module_/dist/pvfs2/2.6.25/pvfs-2.7.1/src/kernel/
>> linux-2.6] Error 2
>> make[1]: *** [default] Error 2
>> make: *** [just_kmod] Error 2
>>

Problem when building 2.7.1 on linux 2.6.25

I have installed the BerkeleyDB and compiled the server and client codes, also 
I have done this "ln /usr/local/BerkeleyDB.4.6/lib/libdb-4.6.so /usr/lib/libdb-
4.6.so".Then I am trying to generate the configuration file on the metadata 
server. However, when I follow the following command it cannot run through.
Like:

PG006:/etc # pvfs2-genconfig /etc/pvfs2-fs.conf /etc/pvfs2-server.conf
Bad arguments.  See -h for help.

But when I change it into this:

PG006:/etc # pvfs2-genconfig /etc/pvfs2-server.conf

it can still run, and then I can continue other successive configs.
But the argument "/etc/pvfs2-fs.conf" is the target where the configurations I 
have done are kept, right? So if I cannot define it, it would be very 
troublesome to install the pvfs2 on a cluster system.

So can anyone help me? I am a novice to pvfs2. Thanks a lot!

Re: Problem when building 2.7.1 on linux 2.6.25

Hi Liu,

In a previous release we changed the configuration file format.  You  
no longer need a per-server config file (/etc/pvfs2-server.conf),  
instead everything gets stored in the global config file (/etc/pvfs2- 
fs.conf).  So when you're running genconfig, only give the first  
argument:

pvfs2-genconfig /etc/pvfs2-fs.conf

The pvfs2-server binary requires only the pvfs2-fs.conf file as an  
argument now as well.

-sam

On May 8, 2008, at 12:55 PM, Liu Yang wrote:

> I have installed the BerkeleyDB and compiled the server and client  
> codes, also
> I have done this "ln /usr/local/BerkeleyDB.4.6/lib/libdb-4.6.so /usr/ 
> lib/libdb-
> 4.6.so".Then I am trying to generate the configuration file on the  
> metadata
> server. However, when I follow the following command it cannot run  
> through.
> Like:
>
> PG006:/etc # pvfs2-genconfig /etc/pvfs2-fs.conf /etc/pvfs2-server.conf
> Bad arguments.  See -h for help.

strange behavior of 2.7.1 client

Hi All,
	Something strange is happening on the client side of 2.7.1. The  
client is running on SLES9 ppc64, and is stuck at traversing a  
particular point in the file system. At the same time accessing other  
parts of the FS work fine, and it appears that servers and the file  
system are working -- pvfs tools read the file system ok, including  
the stuck part. Did anyone see similar behavior?
Thanks,
--andrew

Re: strange behavior of 2.7.1 client

Hi Andrew,

So you're doing a find or some such command and it hangs at a  
particular file?  Is it always the same file?
-sam

On May 9, 2008, at 10:50 AM, Andrew Pochinsky wrote:

> Hi All,
> 	Something strange is happening on the client side of 2.7.1. The  
> client is running on SLES9 ppc64, and is stuck at traversing a  
> particular point in the file system. At the same time accessing  
> other parts of the FS work fine, and it appears that servers and the  
> file system are working -- pvfs tools read the file system ok,  
> including the stuck part. Did anyone see similar behavior?
> Thanks,
> --andrew
>
> _______________________________________________
> Pvfs2-users mailing list
> Pvfs2-users <at> beowulf-underground.org
> http://www.beowulf-underground.org/mailman/listinfo/pvfs2-users

_______________________________________________
Pvfs2-users mailing list
Pvfs2-users <at> beowulf-underground.org

Re: strange behavior of 2.7.1 client

_______________________________________________
Pvfs2-users mailing list
Pvfs2-users <at> beowulf-underground.org
http://www.beowulf-underground.org/mailman/listinfo/pvfs2-users

Re: strange behavior of 2.7.1 client

Sam,
	I also ran pvfs2-fsck on that system (it was still actively used, so  
I don't know how useful the result it.) Here's an output of runnning

# pvfs2-fsck -v /data/P3

Thanks,
--andrew

_______________________________________________
Pvfs2-users mailing list
Pvfs2-users <at> beowulf-underground.org
http://www.beowulf-underground.org/mailman/listinfo/pvfs2-users