added offline server remote deployment config and Dockerfile #7
Conversation
Signed-off-by: Abdul Hameed <ahameed@redhat.com>
| @@ -419,6 +419,14 @@ build-feature-server-java-docker: | |||
| -t $(REGISTRY)/feature-server-java:$(VERSION) \ | |||
| -f java/infra/docker/feature-server/Dockerfile --load . | |||
|
|
|||
| build-offline-server-docker: | |||
There was a problem hiding this comment.
please invoke this target from build-docker target as well
Note Maybe this target is not needed at all, please see next comment
There was a problem hiding this comment.
IIUC, the python feature server already has a Dockerfile that installs the latest feast from pip and a Dockerfile.dev that instead copies the source code in the docker build context and tries to use the latest codebase.
Both run feast serve as the application command through the Helm chart.
Can't we use the same Docker image(s) and just replace the command to feast serve_offline instead? We propably don't need an app.py launcher if the feast application already includes the same with the serve_offline option (and sorry if I gave you wrong pointers before!).
BTW: I just noticed that the dev docker image is affected by this issue #4135, we must involve @jeremyary to discuss a common solution
There was a problem hiding this comment.
we can do that.
if I understand with this idea deploying both servers(online/offline) in a single image within a single pod ? my thoughts was to keep both sever in different pods with it's own image. and Also the offline and online port configuration is different. How the user can configure the port or expose both services externally ?
There was a problem hiding this comment.
In theory we have a fix for the docker version issue, the PR just got stuck in conflict/review limbo.
There was a problem hiding this comment.
@jeremyary what is your thoughts on single image/pod for both offline or online store ?
There was a problem hiding this comment.
@tchughesiv what do you think about having a separate docker image for this server that anyway includes the same application as the python server and runs from a different command like feast serve_offline?
There was a problem hiding this comment.
i was just calling that out that config in case it wasn't known. if you also need an offline deployment config, then helm changes are still the way to go.
There was a problem hiding this comment.
So, a single image is sufficient but we need to fix the issue about dev image
And the offline probably needs a helm condition to run the serve_offline instead of the serve command.
Thanks!
There was a problem hiding this comment.
in helm, you could also do something like feast_mode: offline instead ... which open the door to future ui support?
e.g.
# supported modes - online (default), offline, or ui
feast_mode: offline
There was a problem hiding this comment.
cool, so we can also deploy for free the (undocumented ) registry server that runs as feast serve_registry 😁
Let’s mention this option to Tornike
What this PR does / why we need it:
Added offline server remote deployment config and Dockerfile to generate the offline server image
Which issue(s) this PR fixes:
Fixes