Continuing a discussion that was interrupted by the break, both company and personal :)
Peter Memishian wrote On 12/26/06 11:59 PM,: > > I think what you mean is the zone configuration transition during machine > > upgration. This is similar to other configuration when we upgrade a > machine > > from pre-vanity-naming to post-vanity-naming. We have two choices here: > > > > 1. In the machine migration case, we choose not to give a link a default > > name other than its device instance name (which is bound to the underlying > > hardware. So a hme0 link will still be named as hme0. > > > I agree -- I think we must do (1) on upgrade. So, for systems with > existing zones configured (and thus that are not using vanity names), > the best we can do is walk the zone administrator through some manual > steps that they can use to transition their system -- and thus zones > configured on that system -- to use vanity names. From that point on, > they can make use of vanity names to simplify their zone migrations. > It seems to me that upgrading/transitioning might need a separate section, if not document, because of the unique situation involved. Example, I had a couple of scenarios in mind when thinking of upgrade/transition versus migration. a. Only one system is involved. It has an existing zone configuration (using hme0). That system is then upgraded with Solaris that supports vanity naming. Because it is an upgrade, the link's vanity name will remain hme0. After upgrading, what does the admin do with zone reconfiguration, if any, as far as the network link is concerned, given that there is no name change that actually occurred? b. Two systems are involved. One, without vanity naming support, has an existing zone configuration using hme0. The second system with a ce0 nic, has a freshly installed OS with vanity naming support, and therefore its link is automatically named net0. An admin wishes to migrate the zone to this new system. In normal circumstances, we'd use dladm rename-link to rename net0 to the link that the zone was originally configured to use. However, it would not apply here, since that would mean renaming net0 to hme0 and would counter our intention of weaning link names from their hardware-based names. What does the admin do in this case? Would the only option be a complete revamp of zone configuration in the second system? > As an aside, Raoul: there are other technologies that will be having a > more significant effect on zone administration, such as IP instances and > VNICs. The latter will also make use of vanity naming (since VNICs will > be named by the administrator). To avoid multiple revisions to the zones > documentation that pull it in different directions, I think it's > worthwhile to think about how we want to document the end result after > vanity naming, VNICs, and IP instances are all part of Solaris. > I intend to consult with Steff and the zones writer about how to put all of these things together. Thanks, Raoul
