Hello Orhan,

> So when a node produce a file, the file will be copied to all other nodes, 
> right?

Almost.  All files in the home directories actually reside on the
front-end node;
whenever you try to read or write to a file in the home directory,
this is transparently
routed to the front-end node. (This service and the associated
protocol is called NFS.)

> What if nodes produce a file with same name but different content? Which file 
> will be read by a node?

The last one to write wins.

If two jobs are writing to the same file at the same time, you will
most probably get a corrupted file with some of the data written by
job A and some written by job B, basically at random. (Not really
random, it depends on the exact order the writes arrive to the
server's filesystem.)  This is no different from having two jobs
writing to a local file, though: jobs sharing an output files should
coordinate writes using some mechanism (e.g. locks) otherwise data
corruption occurs.

If you have several jobs writing to the same output files, the safer
thing to do is to start each job in its own directory.

Ciao,
R

-- 
You received this message because you are subscribed to the Google Groups 
"elasticluster" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
For more options, visit https://groups.google.com/d/optout.

Reply via email to