David Eliahu

Persist Prometheus metrics to S3 instead of a persistent volume

1

Thanos can be used for this

Handle AWS termination notice for spot instances

1

#### Motivation Respond to spot instance terminations more gracefully. That is to prevent getting failed requests when the traffic is supposed to migrate from the terminating instance to another one...

Consider adding in-cluster image caching layer

4

The primary motivation for this would be performance and reducing load on an external image registry. One thing to test before investing in it is just how much faster would...

### Description Allow users to specify a list of commands that will be executed during node startup. These will be added to eksctl's `preBootstrapCommands` list.

### Description The prometheus and grafana EBS volumes created by the cluster are not tagged with Cortex's tags (default or user) ### Notes At this time, it does not seem...

#### Description When clustering up, the following warning is emitted: ``` Custom AMI detected for nodegroup cx-ws-cpu-large-2. Please refer to https://github.com/weaveworks/eksctl/issues/3563 for upcoming breaking changes ``` #### Context It seems...

Support websockets in realtime APIs

11

#### Motivation * Reduce latency when multiple requests are required * Stream output from the predictor as it's generated