Thanks Shawn and Erick for explanation... On Thu, Mar 12, 2015 at 9:02 PM, Shawn Heisey <apa...@elyograg.org> wrote:
> On 3/12/2015 9:18 AM, Erick Erickson wrote: > > By and large, I really never use linking. But it's about associating a > > config set > > you've _already_ uploaded with a collection. > > > > So uploading is pushing the configset from your local machine up to > Zookeeper, > > and linking is using that uploaded, named configuration with an > > arbitrary collection. > > > > But usually you just make this association when creating the collection. > > The primary use case that I see for linkconfig is in testing upgrades to > configurations. So let's say you have a production collection that uses > a config that you name fooV1 for foo version 1. You can build a test > collection that uses a config named fooV2, work out all the bugs, and > then when you're ready to deploy it, you can use linkconfig to link your > production collection to fooV2, reload the collection, and you're using > the new config. I haven't discussed here how to handle the situation > where a reindex is required. > > One thing you CAN do is run linkconfig for a collection that doesn't > exist yet, and then you don't need to include collection.configName when > you create the collection, because the link is already present in > zookeeper. I personally don't like doing things this way, but I'm > pretty sure it works. > > Thanks, > Shawn > >