Moving this to the more appropriate common-dev@
I personally like to run a KDC on my dev box and build/deploy a
pseudo-distributed cluster right there. This lets me make principals at
will, adjust krb5 settings as I please, regenerate keytabs, etc. Setting up
a local MIT KDC is actually quite easy.
Aaron T. Myers
Software Engineer, Cloudera
On Thu, Mar 8, 2012 at 4:05 PM, Alejandro Abdelnur <[EMAIL PROTECTED]>wrote:
> Why not just attach a debugger from an IDE (with the right hadoop sources)
> to DN JVM?
> On Thu, Mar 8, 2012 at 1:16 AM, Evert Lammerts <[EMAIL PROTECTED]
> > Hi list,
> > I've spent the whole of yesterday debugging a DN to find and fix a
> > bug (HDFS-3059). Difficulty is that we have Kerberos enabled and I didn't
> > know how to emulate that environment locally. So what I did was editing
> > source, building the jars, deploying them on my NN and a single DN,
> > starting deamons, and looking at my debug statements.
> > Obviously not the most optimal way. How do you debug deamons with
> > enabled? Do you add your development machine to the cluster, give it a
> > principal, and configure the network? Any tips / best practices? I expect
> > we're going to need this more often since we're just bringing up a fair
> > sized multi-tenant production environment (~70 machines) and I guess this
> > won't be the only time we'll want to see what exactly happens.
> > Thanks,
> > Evert