I would vote for this one as well, if somebody wants to create the issue. -----Original Message----- From: "Ted Zlatanov" <t...@lifelogs.com> Sent: Friday, November 20, 2009 3:01pm To: cassandra-dev@incubator.apache.org Subject: Re: overriding directories from command line
On Mon, 16 Nov 2009 09:39:15 -0800 Anthony Molinaro <antho...@alumni.caltech.edu> wrote: AM> If you are thinking about config file changes I'd like to request some AM> sort of inclusion mechanism. For instance, if you could had 2 keyspaces AM> which you want to deploy on separate clusters in production but allow AM> developers to install simultaneously on their development box, it would AM> be great if you could do this by have 3 files. One of which has settings AM> specific to the developer box (ie, memory settings), and one for each AM> keyspace. Then I could create 5 packages (RPM or deb) AM> 1. one with the developer box settings which includes both keyspace configs AM> 2. one with the production box settings for the cluster controlling keyspace1 AM> which includes keyspace1 config. AM> 3. one with the production box settings for the cluster controlling keyspace2 AM> which includes keyspace2 config. AM> 4. one with just the keyspace1 config. AM> 5. one with just the keyspace2 config. AM> I could do all this with a bunch of preprocessing at package build time, AM> but that gets a little annoying. And right now I just have the config AM> copied and everytime I need to change a CF I need to change it in 3 AM> places (devel, qa, and prod configs). AM> One great solution would be to support the loading of multiple configs AM> from a directory. You already see this a lot, for instance /etc/init.d AM> or /etc/httpd/conf.d on most unix systems. This is actually my situation too. I have to roll out slightly customized Cassandra instances to several colos and it's unpleasant to use cfengine or Puppet to edit XML files, so I'd rather have some simple inclusion and interpolation mechanism. Ted
