If the absolute filepath is the same in the machines (since SHA-1 hash of filepath is used as engineVersion), there would be no changes in the above workflow. You can just copy the engine directory and deploy. If not, you can also provide engine-id, engine-version as CLI arguments.
There's some additional ongoing work to make the process more robust and clearer, such as enabling engine-directory as CLI input, help messages, and adding tests to cover such scenarios. I don't think the PR will be merged before then. On Mon, Dec 12, 2016 at 11:49 AM, pferrel <[email protected]> wrote: > Github user pferrel commented on the issue: > > https://github.com/apache/incubator-predictionio/pull/328 > > too much noise in @dev, I have a simple question. > > If I train on one machine, how do I deploy on another. Currently I > copy the manifest--well actually the entire directory--from the train > machine to the deploy machine. > > This is actually a very important use case because the train machine > has the Spark driver on it so it may have to be very large (and temporary, > but the deploy machine can be smaller and far cheaper. Not to mention > having multiple PredictionServers behind a load balancer. > > > --- > If your project is set up for it, you can reply to this email and have your > reply appear on GitHub as well. If your project does not have this feature > enabled and wishes so, or if the feature is enabled but not working, please > contact infrastructure at [email protected] or file a JIRA ticket > with INFRA. > --- >
