Re: Rookie Error?

  • From: Matthew Zito <mzito@xxxxxxxxxxx>
  • To: Rich.Jesse@xxxxxxxxxxxxxxxxx, Oracle-L Freelists <oracle-l@xxxxxxxxxxxxx>
  • Date: Mon, 29 Nov 2004 11:55:39 -0500

Yeah, in a RAC environment what makes it angry is trying to call 
interface status ioctls on the loopback interface - which of course 
isn't a real interface at all.  Programatically, there's no reason 
Oracle couldn't have had it return some sane error message like, "Tried 
to attach to interface that does not support necessary information - ip 
(blah) interface (blah)", etc. etc.  Instead, the cluster manager just 
crashes with gibberish and numbers that obviously indicate type 

10g is extra dumb, because it actually uses the output from uname() or 
gethostname() (we haven't determined which) to determine its node 
identity in a CRS config - its not smart enough to store ip 
information, or be told that on the fly.  When we implemented 10g 
support in our software, that was one of the multitude of new 10g RAC 


Matthew Zito
GridApp Systems
Email: mzito@xxxxxxxxxxx
Cell: 646-220-3551
Phone: 212-358-8211 x 359

On Nov 29, 2004, at 11:27 AM, Jesse, Rich wrote:

> Eeep!  Same thing burned me when setting up RAC, but it was only the 
> private net that was hosed (hadn't gotten as far as attempting a 
> connection).  I only found it when I retraced every stinking setup 
> step-by-step and noticed that my /etc/hosts was just that little 
> different from the example RAC setup.
> Rich
> -----Original Message-----
> Sent: Sunday, November 28, 2004 4:43 PM
> To: mzito@xxxxxxxxxxx
> Cc: Oracle-L Freelists
> Subject: Re: Rookie Error?
> Thanks Matt, no worries, I understood the first time.
> That was a masterful bit of sleuthing Matt, you nailed it.
> The IP for sherlock was set to in sherlock:/etc/hosts.
> Reset it to, restart the listener, all is well.
> Thanks, I hope this has been burned into my memory now.  :)
> Jared
> --


Other related posts: