Question: Is anyone loadbalancing their deployment

[kriswest]

Is anyone currently loadbalancing a deployment of GIt Proxy with multiple nodes? I believe the app should loadbalance fine when deployed using MongoDB - we're currently deployed with an LB, but only one node, but are looking to switch to a 2 node arrangement and try to achieve zero downtime during deployments, upgrades etc.

It would be great to include a note on this in the deployment guide @tabathad is contributing in #1385

[coopernetes]

(converted into a discussion in case other deployment ideas are worth sharing)

Our deployment at @RBC is multi-instance. We're running multiple pods of GitProxy in a Kubernetes environment and load balancing using standard Ingress/Service objects. MongoDB is external to the app so persistence is durable across restarts. I haven't seen any real sharp edges with this deployment. Git over HTTPS is kind in that it is only a few actual TCP connections and HTTP requests so it can be load balanced nicely. The protocol being stateless doesn't hurt either.

On v1, we did have to modify some of the Vite build setup in order to properly pass in dynamic URLs for the frontend (which we eventually disabled anyways) to properly respond on Ingress hosts across different environments/clusters. This is required due to CORS. I think the latest changes we made slated for v2 (via #1165) should make setting the base URL for the UI simpler via environment variables.