No dia 8 de Outubro de 2010 07:28, Fechner, Jon L. <[email protected]> escreveu: > Just a note to let you know we do share over NFS and are benefiting from it. > We have one installation for Solaris 8, 9, and 10 (on sparc) and it allows us > to easily keep all users on the same version of the tools. > > We do not plan on abandoning this approach so I hope support is continued. > > I would be interested to know what the significant compromises are in package > delivery if you have time, but it is not pressing.
Here's one example. Suppose that you deliver configuration files to /etc/opt/csw so that they are zone-local on sparse zones with shared /opt/csw. If you do that, they configuration files are not available on the shared NFS solution, because they aren't under the shared /opt/csw directory. If you choose to deliver configuration files to /opt/csw/etc, they are available, if not modifiable, in the shared NFS solution. The downside is that you cannot modify then on sparse non-global zones. My position on the topic is that if you're using the shared NFS solution, you can use a configuration management system such as puppet or cfengine to deploy your configuration files to /etc/opt/csw. I would assume that any larger Solaris installation requires such a solution, but I heard that there is a deployment that does not use any configuration management system, and it requires configuration files under /opt/csw/etc. There are also issues with alternatives, and generally class action scripts. How are NFS-based deployments managing the configuration files, do you guys use a configuration management system, or is it purely NFS-shared /opt? _______________________________________________ users mailing list [email protected] https://lists.opencsw.org/mailman/listinfo/users
