[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Condor-users] newbie - condor remote job submission problem



On Mon, Feb 21, 2005 at 05:31:54PM +0530, krishnaprasad wrote:
> 
> 
> Hai all
> 
> Iam facing a problem in remote job submission. If Iam trying to do
>    
>                 condor_submit -r <machine name>  jobfile
> 
> then the following error is coming: -
> 
>     ERROR : Failed to connect to queue manager <machine name>
>     AUTHENTICATE : 1003: Failed toauthenticate with any method
>     [ more errors ]

submitting a job requires that you authenticate yourself to the schedd.
normally on a unix machine, when you submit to a schedd running on your
local machine, this is accomplished using 'FS' authentication, which means
filesystem.  the condor_submit process writes a file in /tmp, and the schedd
looks at the owner of that file to see if you are who you say you are.  now,
this method will obviously fail when submitting remotely because each machine
has it's own /tmp.

there are a number of things you can do:
1) if you require real security, you can use KERBEROS (difficult to set up
if you aren't already using it) or GSI which is slightly easier to set up.

2) you can essentially disable the authentication by using the authentication
method 'CLAIMTOBE'.  this isn't included in the default list because it is not
secure, but it will always succeed.  to do this, add a line like this to your
condor_config:
  SEC_DEFAULT_AUTHENTICATION_METHODS = CLAIMTOBE

3) it is possible, though not recommended at all, to use a shared filesystem
like NFS for authentication.  then, instead of writing into /tmp, you can
choose your own directory for condor to use for authentication.  if you then 
choose a shared directory that both the submit machine and the machine with
the schedd can access, you can use FS_REMOTE.  the downside of this is that
it sometimes fails (especially under load) because the two sides do not sync
properly.  this will be addressed later in the 6.7.X series but for now the
FS_REMOTE method should only be used for experimentation.

let me know if i can help answer more questions.  for more info you can also
read the security section of the manual here:
  http://www.cs.wisc.edu/condor/manual/v6.6.8/3_7Security_In.html


cheers,
-zach