Hi Fedor, > Use 'node-{ID}-{#}' format, where {#} we'll chose in loop till the first > unique.
I don't like this approach by many reasons. Here's some of them: * With a loop you're going to perform N SQL queries in order to check for uniqueness, and that's a bad design and it'd be better to avoid it. * What if user wants to use exactly the same schema? For instance, he use the first number as a rack id and the second one as a node id in this rack. The proposed approach will implicitly set some wrong hostname, and detecting it may be a challenge. > Use some unique value, shorter than UUID (for example - number of > microseconds from current timestamp) Usually, it's incredibly rare situation, so solving it by adding UUID is a more than enough. Moreover, it will take attention of the cloud operator, so he will easily detect a conflicting node and fix its name as he wishes. And that's probable what he wants to do almost all the time - to fix its hostname, none want to work with names like: node-1, node-2, node-2-2, node-3, because what is the node-2-2? Does it store some backup or what? It confuses. So, I think current approach - node-UUID - is the way we should do. Thanks, Igor On Tue, Jul 21, 2015 at 3:38 PM, Fedor Zhadaev <fzhad...@mirantis.com> wrote: > Hi all, > > The next issue was found during implementation > https://blueprints.launchpad.net/fuel/+spec/node-naming : > > User may change node hostname to any another, including default-like > 'node-{№}', where № may be bigger than maximum nodeID existing at that > moment. > Later when node with ID == № will be created it's default name 'node-{ID}' > will break hostnames uniqueness. > > To avoid this now it was decided to generate in such situation another > default hostname. > > The current solution is to generate hostname 'node-{UUID}'. It works, but > may look terribly. > > There are a few another possible solutions: > > Use 'node-{ID}-{#}' format, where {#} we'll chose in loop till the first > unique. > Use some unique value, shorter than UUID (for example - number of > microseconds from current timestamp) > > Please share you opinion - what is better? > > Also you can propose your own solutions. > > -- > Kind Regards, > Fedor Zhadaev > Junior Software Engineer, Mirantis Inc. > Skype: zhadaevfm > E-mail: fzhad...@mirantis.com > > __________________________________________________________________________ > OpenStack Development Mailing List (not for usage questions) > Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe > http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev > __________________________________________________________________________ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev