E
Eggshell Tyrannosaurus
Impact: Teams would no longer have a way to validate their applications post-deployment.
Workaround: There are two.
  • The first is to deploy Harness so that it has direct access to the clusters. We can either do this with or without ArgoCD but the system would have a great deal of access to our clusters and it would potentially dramatically change our cloud platform build process.
  • The second is to maintain the EventRouter component and have that system call back into Harness to invoke a validation pipeline after it sees ArgoCD’s success. This requires us to need to maintain our current code.
If we use the system where Harness does not have direct control of the clusters, which matches better with our current architecture, Harness would not immediately have a way to know when a deploy is successful. This is because it would perform a gitops operation (merge a PR) and then it would need to wait for a signal that the operation had completed so that it knew to check to see if the deploy had been successful. Since it doesn’t have access to the cluster there’s no direct interface to know this. Harness would have to implement a way to get a signal back from ArgoCD to trigger it to run the verification, but without having direct access to the cluster.
Created by Dominique Feldman
·