Hey everyone! I'm really frustrated and need to know if I'm the only one dealing with this bizarre issue between ArgoCD and Crossplane. Despite ArgoCD showing resources as "Healthy" and "Synced", I'm encountering serious failures with Crossplane when trying to provision AWS resources. I'm getting repeated 400 errors from AWS while ArgoCD remains completely oblivious, showing everything as just fine, which is super misleading. I've been googling for days and can't find anything related to this. What's happening is that the health check logic in Lua is deciding that if `Ready: True` appears before `Synced: False`, ArgoCD believes everything is healthy, ignoring real failures. I can't believe nobody else seems to have hit this. Are you all just avoiding health checks or monitoring through AWS directly? Or am I just uniquely unlucky? I managed to resolve it by changing the condition checks, but I truly expected more awareness in the community about this problem. Any insights?
5 Answers
Honestly, I think there’s a fundamental misunderstanding here. The resources are actually synced, but if Crossplane fails afterwards, that’s independent of Argo's checks. GitOps is more about ensuring your cluster resources match the desired state, not guaranteeing everything's 'up' constantly. It’s about having proper monitoring set up for health checks. You might want to explore using a tool like Grafana or Prometheus for better insights into your system's health.
For sure. Lies in the integration, not ArgoCD’s responsibility.
Thanks for sharing your experience! I’m considering a similar stack migration and really appreciate the insights. Have you thought about documenting this as a GitHub issue? It sounds like it might help others who face the same challenge.
I considered that, but it seemed like the maintainers were already prioritizing other issues. Writing up the article felt more beneficial for now!
Awesome that you found a workaround! But seriously, Medium's member-only stories can be a bit frustrating. Maybe consider sharing your solution somewhere more accessible?
Same here, if it’s gated like that, I tend to steer clear!
Exactly! I usually skip over Medium unless it's open access. Just doesn't sit right.
I've been in the same boat, but I learned about Argo's health check behavior beforehand. A good practice is to create custom health checks for your resources, especially if the defaults don’t work well with your setup.
I still think more people should know about this. It shouldn’t be that hard to learn!
Why not report a GitHub issue directly? It could potentially lead to a fix instead of just sharing a story on Medium. It’s worth a shot!
I heard from maintainers that they’re not treating it as a major issue, which is frustrating.
Totally agree, ArgoCD isn't designed to be a health dashboard. It’s more about deployment management. If Crossplane is failing, it should be caught with logs or metrics instead.