[Cado-nfs-discuss] Fwd: Re: Lingen solution not found

Emmanuel Thomé emmanuel.thome at gmail.com
Sun Jul 13 13:58:46 CEST 2014


The list should have the reply to list set up...
---------- Message transféré ----------
De : "Emmanuel Thomé" <emmanuel.thome at gmail.com>
Date : 13 juil. 2014 13:56
Objet : Re: [Cado-nfs-discuss] Lingen solution not found
À : "Soundcheck Team" <soundcheck2014 at gmail.com>
Cc :

Hi.
>
> Files can be corrupted before the point where they get created with size
> zero. Some earlier files may have been created with holes due to failed
> writes. This translates to long blocks of zero bytes within the files. Or
> the files may be shorter than expected.
>
> Note also that the V files are used for recovery and may have been
> affected similarly by disk full conditions.
>
> Verifying krylov data post-mortem is doable, but the functionality is
> lacking presently in cadonfs. If you want to avoid headaches, the most
> robust way is to not count on recovery as long as you fear that corrupted
> files could remain.
>
> Best regards,
>
> E.
> Le 13 juil. 2014 11:26, "Soundcheck Team" <soundcheck2014 at gmail.com> a
> écrit :
>
>> Hi,
>> Thanks for your helpful response!
>> We are using version 2.0 updated to commit
>> 39075958e0fda9e3187f60719bc61f3c96d206ac (The last build about when we
>> started).
>> We did get a full disk in the middle of the krylov phase, which caused A*
>> files (files with names like A0-64.0-149900) to be created with length 0.
>> We stopped the process and erased all files in the c204.bwc directory
>> created since the first such file was created, ghosted the disk to a larger
>> one and then resumed. Could there still be corrupted files? We also had to
>> resume a few times due to server crashes.
>> I'm attaching the last bwc logs; However the whole log file is 486 MB; is
>> there a specific part of it that could help?
>> Thanks again,
>> Eran
>>
>>
>> On 9 July 2014 21:45, Zimmermann Paul <Paul.Zimmermann at inria.fr> wrote:
>>
>>>        Dear Eran,
>>>
>>> > Date: Wed, 9 Jul 2014 19:59:00 +0300
>>> > From: New Name <soundcheck2014 at gmail.com>
>>> >
>>> > Hi,
>>> > We've been using CADO-NFS to factor a large number (640 bit), finally
>>> > reached the lingen subphase of the Linear Algebra phase but we got an
>>> error
>>> > - "no solution found". Are there any known issues which could cause
>>> this?
>>> > Notable issues in our run so far:
>>> > - We're using the default c204 params file even though the number is a
>>> bit
>>> > smaller (196 digits) as there is no 196 digit param file.
>>>
>>> this should not be a problem.
>>>
>>> > - Are there any issues in lingen when lpbr>32 ? Should we have used non
>>> > default parameters?
>>>
>>> no, we already used a large prime bound of 2^33 for RSA-704 for example.
>>>
>>> > - We had to resume the krylov sub phase a few times.
>>>
>>> my guess is that one of the checkpoints was corrupted. Usually this is
>>> due to a full disk.
>>>
>>> > Any help would be wonderful,
>>> > Eran
>>>
>>> please can you send us the full *.log file, and the *bwc*stdout and
>>> *bwc*stderr files? Also which version of CADO-NFS are you using?
>>>
>>> Best regards,
>>> Paul Zimmermann
>>>
>>>
>>
>> _______________________________________________
>> Cado-nfs-discuss mailing list
>> Cado-nfs-discuss at lists.gforge.inria.fr
>> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/cado-nfs-discuss
>>
>>
-------------- section suivante --------------
Une pi�ce jointe HTML a �t� nettoy�e...
URL: <http://lists.gforge.inria.fr/pipermail/cado-nfs-discuss/attachments/20140713/d4dbef56/attachment.html>


More information about the Cado-nfs-discuss mailing list