substratusai / lingo

Lightweight ML model proxy and autoscaler for kubernetes
https://www.substratus.ai
Apache License 2.0
102 stars 6 forks source link

Redeploying lingo causes deployments to get lost #30

Closed samos123 closed 7 months ago

samos123 commented 8 months ago

Steps to reproduce:

  1. helm install lingo substratusai/lingo
  2. Install a deployment to be used with lingo:
    helm upgrade --install stapi-minilm-l6-v2 substratusai/stapi -f - << EOF
    model: all-MiniLM-L6-v2
    replicaCount: 0
    deploymentAnnotations:
    lingo.substratus.ai/models: text-embedding-ada-002
    EOF
  3. Re-install lingo: helm uninstall lingo && helm install lingo substratusai/lingo

Current result: Lingo will no longer be able to list and scale stapi deployment even though it's still there

Expected result: Lingo should still work on all existing deployments even after a re-install

nstogner commented 8 months ago

I thought I reproduced this by watching the logs and not initially seeing mention of stapi. However after starting a new port-forward and sending a new request, lingo scaled up the stapi deployment after re-install of lingo.

samos123 commented 7 months ago

I can't reproduce either. Closing