chatbot-rag-app: adds Kubernetes manifest and instructions #396
Conversation
codefromthecrypt
requested review from
xrmx,
trentm,
joemcelroy,
estolfo,
bshetti,
stanek-michal and
EvelienSchellekens
|
note: each thing we do runs back into this. it would be great to have a way to quickly initialize elser not just installing it, but first time use without timeouts for several minutes #307 |
|
I have work almost done to make this "normal k8s" local, but wanted to solve the timeout first. so I'll push commit after #397 is merged |
|
will bump this tomorrow or when an approver looks at #397 |
codefromthecrypt
force-pushed
the
k8s-chatbot-rag-app
branch
from
9f0bb96 to
3660c11
Compare
|
rebased and changed to non-host network k8s. will leave this in draft until #397 is merged as using not-yet-deployed images in k8s is a pain. |
|
waiting to get the docker image smaller before "ready for review", as I noticed my network lagging #407 |
Signed-off-by: Adrian Cole <[email protected]>
codefromthecrypt
force-pushed
the
k8s-chatbot-rag-app
branch
from
3660c11 to
5145b72
Compare
Signed-off-by: Adrian Cole <[email protected]>
|
ok things work in general, but I'm not seeing traces in kibana. I have to put this down for a bit as I have other more urgent things to address. |
|
|
||
| Note: If you haven't checked out this repository, all you need is one file: | ||
| ```bash | ||
| wget https://raw.githubusercontent.com/elastic/elasticsearch-labs/refs/heads/main/docker/docker-compose-elastic.yml |
There was a problem hiding this comment.
Think this is wrong file
| ports: | ||
| - containerPort: 8200 | ||
| command: | ||
| - /usr/local/bin/docker-entrypoint |
There was a problem hiding this comment.
TBH didn't go so far as running chatbot yet but could confirm traces for normal openai examples with this change
There was a problem hiding this comment.
good eye!
| labels: | ||
| app: chatbot-rag-app | ||
| spec: | ||
| # The below will recreate your secret based on the gcloud credentials file |
There was a problem hiding this comment.
I personally find the GCP-specific instructions in the readme and manifest to be sort of jarring and one point is this is mostly for running locally, while in a cluster the pod should use service account auth provided by the cluster. Similar can be said about using static AWS credentials via env.
"Note for cloud provided backends such as Vertex AI and Bedrock, the pod will need to be authenticated with access to them. This is usually with service account authentication in running clusters, while if testing locally can be with AWS static credentials passed by environment variables or copying a GCP credentials file into a secret mounted as a volume."
Not sure if this is too much information though but I think it would be good if this yaml could avoid the gcloud-specific parts.
There was a problem hiding this comment.
good point, I also was troubled by it. better to leave this out as it is arduous and creates FUD.
|
Due to elasticon singapore and Sydney... while excited about this i am not finishing it this weekend. Maybe Tuesday |
Decided to action this so that we have a coherent experience between docker compose and k8s. This is as close as I could get it. If folks have feedback or a different direction, do tell!
Fixes #366