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

Re: [Condor-users] Will HAD_CONTROLLEE = SCHEDD can be used for condor_schedd HA .



> Hi Nick,
Hello Johnson,
 
>     So now we don't need SCHEDD.lock at spool directory is it am I right.
> 
>     Instead of keeping $(SPOOL) as shared. Why can't condor_replication
> functionality is used for that keeping the directory in sync. so that
> dependency from the shared storage will be removed (i.e nfs etc)

The current replication software only allows for the replication of a single 
file.  I have updates to it that I need to finish testing and merge into the 
master branch that would allow for this.  However, even this update doesn't 
allow for replicating all of the job specific files that come and go into the 
spool directory.  AKA you can't just tell it to "replicate spool and all of 
the files and subdirectories in it."  I hope to get this work into the 7.5.x 
series.

If that's what you really want, perhaps you could do something with running 
rsync periodically.

FWIW, we had discussed replacing the core of the replicator with rsync, but 
decided against that because rsync doesn't run on Windows (well, without 
something like cygwin).

I need to think a bit about this...  With the restructuring of the replictor 
that's gone into my branch, perhaps there would be a way to wedge in an rsync 
in place of the transferer...

In any case, with the current 7.4.x and 7.5.x software doesn't allow you to do 
what you're asking...

-Nick

-- 
           <<< The Matrix is everywhere. >>>
 /`-_    Nicholas R. LeRoy               The Condor Project
{     }/ http://www.cs.wisc.edu/~nleroy  http://www.cs.wisc.edu/condor
 \    /  nleroy@xxxxxxxxxxx              The University of Wisconsin
 |_*_|   608-265-5761                    Department of Computer Sciences