[Date Prev][Date Next] [Chronological] [Thread] [Top]

Re: distributed directories



We're essentially doing the same thing, but without the CVS (overkill for our 
purposes) or krsh.

To be honest the very thought of kerberized rsh gives me hives!  :) 

But yah, test test test is the mantra for us too.

--Charlie

On 14 Apr 2004 at 11:39, Quanah Gibson-Mount wrote:
> 
> 
> --On Wednesday, April 14, 2004 11:19 AM -0700 Donn Cave 
> <donn@u.washington.edu> wrote:
> 
> 
> There are really all sorts of ways to do this.  We keep two conf files, a 
> conf for the master, and a conf for the slaves.  The conf for the slaves is 
> essentially identical bar hostnames.  So we use the tool called "bundle" 
> (<http://www.eyrie.org/~eagle/software/bundle/>) to filter in the 
> hostnames, and use krsh (kerberized rsh) to distribute updates to the conf 
> files on all our servers.  All changes to the conf files are also tracked 
> in CVS, so we can back out if there are issues later on.  Of course, we 
> test conf file changes on our test servers before deploying them to 
> production. ;)
> 
> --Quanah
> 
> 
> --
> Quanah Gibson-Mount
> Principal Software Developer
> ITSS/TSS/Computing Systems
> ITSS/TSS/Infrastructure Operations
> Stanford University
> GnuPG Public Key: http://www.stanford.edu/~quanah/pgp.html