[gridway-user] gridway dicovers hosts only
Ruben Santiago Montero
rubensm at dacya.ucm.es
Thu Sep 6 10:33:17 CDT 2007
Hi Jim
It seems that the user mappings are somehow wrong. Let see thefollowing
configuration does not work:
host1.
Running gwd
User jim has a valid proxy there.
host2
DN "jim" is mapped to local user jim.
gridway can not to create directories and transfer files.
right?
Are host1 and host2 configure to use NIS or NFS?. or do they share by other
means the user namespace and home filesystem?
Regards
Ruben
On Tuesday 04 September 2007 14:21, Jim Davies wrote:
> Ruben,
>
> Thank you for your e-mail.
>
> I ran the gw_tm_mad_ftp command as suggested for user 'jim', it worked
> fine - I copied /etc/passwd to the remote server, location /tmp/pass
> no problem.
>
> I resolved the remote directory creation problems for
> gsiftp://host1.grid/~/.gw_gwadmin_3/, by setting up globus security
> for the remote gwadmin user (my oversight). This solves job submission
> by user gwadmin, which is now all OK.
>
> But ..... the job.log files for jobs submitted by 'jim' still
> reference the gwadmin location 'gsiftp://host2.dn/~/.gw_jim_1/' ('~'
> being shorthand for 'gwadmin'). I guess user 'jim' does not have
> sufficient rights to the 'gwadmin' home and this is the reason for the
> failure.
>
> The job gets as far as epilog, but then just loops.
>
> Please see the job.log output and job.contact output below:
>
> job.log
> =======
> Tue Sep 4 15:14:04 2007 [TM][I]: Staging output files:
> Tue Sep 4 15:14:04 2007 [TM][I]: Source:
> gsiftp://host2.dn/~/.gw_jim_1/.
> Tue Sep 4 15:14:04 2007 [TM][I]: Copying file stdout.wrapper.
> Tue Sep 4 15:14:04 2007 [TM][I]: Copying file stderr.wrapper.
> Tue Sep 4 15:14:05 2007 [TM][I]: File stdout.wrapper copied.
> Tue Sep 4 15:14:06 2007 [TM][I]: File stderr.wrapper copied.
> Tue Sep 4 15:14:06 2007 [TM][I]: All output files copied.
> Tue Sep 4 15:14:06 2007 [DM][I]: New state is EPILOG.
> Tue Sep 4 15:14:06 2007 [TM][I]: Staging output files:
> Tue Sep 4 15:14:06 2007 [TM][I]: Source:
> gsiftp://host2.dn/~/.gw_jim_1/.
> Tue Sep 4 15:14:06 2007 [TM][I]: Copying file stdout.execution.
> Tue Sep 4 15:14:06 2007 [TM][I]: Copying file stderr.execution.
> Tue Sep 4 15:14:06 2007 [TM][I]: Retrying copy of file
> stdout.execution in ~5 seconds.
> Tue Sep 4 15:14:06 2007 [TM][I]: Retrying copy of file
> stderr.execution in ~5 seconds.
> Tue Sep 4 15:14:07 2007 [TM][I]: Retrying copy of file
> stdout.execution in ~10 seconds.
> Tue Sep 4 15:14:07 2007 [TM][I]: Retrying copy of file
> stderr.execution in ~10 seconds.
> Tue Sep 4 15:14:17 2007 [TM][E]: Copy of file stdout.execution
> failed. Tue Sep 4 15:14:17 2007 [TM][E]: Copy of file
> stderr.execution failed. Tue Sep 4 15:14:17 2007 [TM][W]: Some output
> files were not copied, will NOT remove remote directory.
> Tue Sep 4 15:14:17 2007 [DM][E]: Epilog failed:
>
> job.contact
> ===========
> https://172.70.12.140:8443/wsrf/services/ManagedExecutableJobService?6dd5f9
>70-5af1-11dc-9ca8-952be0d26d88
>
> Quoting Ruben Santiago Montero <rubensm at dacya.ucm.es>:
> > Hi Jim,
> > Can you transfer files from/to host2.dn using the globus commands? The
> > problem is that the Transfer Manager is not able to create a directory in
> > host2.dn. If you are able to use the GridFTP server at host2.dn, check
> > that the TM is working:
> >
> > 1.- Execute the MAD (if you are using a multipleuser installation, do it
> > through sudo,
> > sudo -u <user> $GW_LOCATION/bin/gw_tm_mad_ftp)
> >
> > 2.- type in the following (lines starting with '>'
> > are responses to the commands):
> > INIT 200 - - - -
> >
> > >INIT - - SUCCESS -
> >
> > START 0 - - - -
> >
> > >START 0 - SUCCESS -
> >
> > CP 0 0 - file:/etc/passwd gsiftp://host2.dn/tmp/pass
> >
> > >CP 0 0 SUCCESS (file:/etc/passwd->gsiftp://host2.dn/tmp/pass)
> >
> > END 0 - - - -
> >
> > >END 0 - SUCCESS -
> >
> > FINALIZE - - - - -
> >
> > >FINALIZE - - SUCCESS -
> >
> > Let see if that works
> >
> > Best
> > Ruben
> >
> > On Friday 31 August 2007 15:36, you wrote:
> >> Eduardo,
> >>
> >> Thanks for the reply.
> >>
> >> I seem to have fixed my problem (at least the one re. passing MDS data
> >> to Gridway via IM_MAD).
> >>
> >> I reviewed the output produced by the wsrf-query below:
> >>
> >> wsrf-query -s https://host1.dn/wsrf/services/DefaultIndexService '/*'
> >>
> >> ... and noticed that it listed sections for Condor that had unknown or
> >> NULL fields, I think it was these that were being picked up by
> >> Gridway. I moved a few superfluous directories e.g.
> >> gram-service-Condor and gram-service-PBS which has sorted things out...
> >>
> >> The great advance is that gwhost now lists correctly, jobs are being
> >> submitted and some actioned, but....looking at the job.log for a job
> >> running 'ls -al' on a remote node, I noticed the job.log error:
> >>
> >> Fri Aug 31 16:24:00 2007 [DM][I]: New state is PENDING.
> >> Fri Aug 31 16:24:48 2007 [DM][I]: New state is PROLOG.
> >> Fri Aug 31 16:24:48 2007 [TM][I]: Creating remote job working directory:
> >> Fri Aug 31 16:24:48 2007 [TM][I]: Target url:
> >> gsiftp://host2.dn/~/.gw_jim_2/.
> >> Fri Aug 31 16:24:48 2007 [TM][E]: Could not create remote job
> >> directory (gsiftp://host2.dn/~/.gw_jim_2/).
> >> Fri Aug 31 16:24:48 2007 [DM][I]: Prolog failed:
> >>
> >> What would cause this?
> >>
> >> Jim
> >>
> >> Quoting Eduardo Huedo Cuesta <ehuedo at fdi.ucm.es>:
> >> > Hi Jim,
> >> >
> >> > What does the IM MAD answer when you execute it directly and type
> >> > "MONITOR 0 host1.dn"?
> >> > Also, can you send us the gwd.log?
> >> > There is a "::" in your IM MAD configuration which is just ":".
> >> > Anyway, I think it is not related to the current problem (but it could
> >> > be the next one).
> >> >
> >> > Regards,
> >> >
> >> > Eduardo.
> >> >
> >> > Jim Davies escribió:
> >> >> Hi,
> >> >>
> >> >> I have globus-4.0.5 and Gridway configured and have tested gwsubmit
> >> >> using the following 'static' IM MAD setting in gwd.conf: IM_MAD =
> >> >> mds4:gw_im_mad_static: -l
> >> >> usr/local/globus-4.0.5/etc/gridway/im_examples/host.list:gridftp:ws,
> >> >> which works just fine.
> >> >>
> >> >> When I switch to a 'dynamic' setting using: IM_MAD =
> >> >> mds4:gw_im_mad_mds4:-s host1.dn::gridftp:ws, gwd.log only reports
> >> >> hostname discovery and gwps lists the job as pending.
> >> >>
> >> >> I have run gw_im_mad_mds4 -s host1.dn from the command line,
> >> >> entered DISCOVER and had a couple of hostnames returned only, which
> >> >> fits with what I'm seeing in the log.
> >> >>
> >> >> *I have ganglia running and the output of wsrf-query on the
> >> >> DefaultIndexService suggests that ganglia is feeding MDS. Also, a
> >> >> telnet to localhost port 8649 returns expected values.
> >> >>
> >> >> *I have also setup server-config.wsdd as recommended and pointed it
> >> >> to the gluece-rpprovider-sample-config.xml and visited
> >> >> http://www.globus.org/toolkit/docs/4.0/info/usefulrp/
> >> >>
> >> >> I would be grateful for any further advice on how to get around
> >> >> this issue or where I'm going wrong.
> >> >>
> >> >> Rgds,
> >> >>
> >> >> Jim
> >> >
> >> > --
> >> >
> >> > **************************************************
> >> >
> >> > Dr. Eduardo Huedo Cuesta
> >> >
> >> > Associate Professor (Prof. Contratado Doctor)
> >> > Dpto. Arquitectura de Computadores y Automática
> >> > Facultad de Informática
> >> > Universidad Complutense de Madrid
> >> > Spain
> >> >
> >> > http://asds.dacya.ucm.es/ehuedo
> >> >
> >> > **************************************************
> >
> > --
> > +-----------------------------------------------------------+
> > Dr. Ruben Santiago Montero
> > Associate Professor
> > Dpto. Arquitectura de Computadores y Automatica
> > Facultad de Informatica
> > Universidad Complutense phone : +34 91 394 75 38
> > 28040 Madrid fax : +34 91 394 75 27
> > Spain email : rubensm at dacya.ucm.es
> > http://asds.dacya.ucm.es/
> > +-----------------------------------------------------------+
> >
> > GridWay, The Way to Grid! http://www.gridway.org
--
+-----------------------------------------------------------+
Dr. Ruben Santiago Montero
Associate Professor
Dpto. Arquitectura de Computadores y Automatica
Facultad de Informatica
Universidad Complutense phone : +34 91 394 75 38
28040 Madrid fax : +34 91 394 75 27
Spain email : rubensm at dacya.ucm.es
http://asds.dacya.ucm.es/
+-----------------------------------------------------------+
GridWay, The Way to Grid! http://www.gridway.org
More information about the gridway-user
mailing list