This folder contains sample applications for Vespa deployments in different topologies, environments, services and platforms.
Vespa users run on a wide range of stacks - we welcome contributions once you got your app running!
Use the AWS multinode quick-start guides in getting started.
See vespa.ai/support for how to get help / FAQ / Stack Overflow. Admin procedures can be a good read, too.
The Multinode application sets up a three-node application and experiment with node start/stop. Use status and metrics pages to inspect the system. Whenever deploying and facing issues, please refer to this application for how to get useful debugging information for support follow-up - i.e. run this application first.
The Multinode High Availability application is a great starting point for high-availability multi-node deployment of Vespa. The example deploys 10 nodes on one base host, and is a good example of a full-fledged multinode application - use this as a template. This example includes securing the application using mTLS.
The Vespa on Kubernetes (Google Kubernetes Engine) application demonstrates deploying a simple Vespa application on Kubernetes (Google GKE).
These are monitoring and metrics-oriented applications.
The Vespa metrics and monitoring application demonstrates how to integrate Vespa with Prometheus and Grafana.
Developers have reported problems running a multinode Vespa cluster on Kubernetes.
Root cause is often the three Vespa Config Servers that uses a shared ZooKeeper cluster,
see multinode for details.
The Zookeeper cluster must be up before all Config Servers are up on http://host:19071/ApplicationStatus.
If 19071 is used in the readinessProbe, the Kubernetes service does not register the host, and there is a catch-22.
Removing the readinessProbe on 19071 can make the Config Server cluster start /
set publishNotReadyAddress: true - also see basic-search-on-gke and
vespa-quick-start-kubernetes.