I've been stuck trying to figure out a major issue with ArgoCD and Crossplane that's driving me bonkers! ArgoCD shows resources as 'Healthy' and 'Synced,' yet Crossplane is struggling to provision AWS resources, returning 400 errors. It feels like I'm the only one dealing with this, especially since I've found no solid documentation or discussions about it. The Lua logic for health checks seems flawed – it processes conditions in an array, meaning that if a resource is marked as 'Ready: True' before 'Synced: False,' ArgoCD just assumes everything is fine, ignoring the reality of failing cloud resources. Is it just me facing this? Are others out there monitoring their systems differently? I kind of fixed it by adjusting the order of condition checks, but I can't believe this isn't more widely known in the community! Please tell me I'm not alone in this!
5 Answers
I think there's a misunderstanding here regarding GitOps and ArgoCD. ArgoCD is doing its job; it reflects the state of the resources as they're synced. The failures you're facing afterward don't invalidate its assertions. Remember, GitOps ensures the desired state matches what's in the cluster—it doesn't equate to having everything healthy and running perfectly. You should set up proper monitoring to catch those failures.
Exactly! Argo is meant for deployment, not health assessment. Try integrating some observability tools.
I've dealt with this issue too! Thankfully, I knew about Argo's health check quirks already. I assumed most users would test and write custom health checks for their resources to avoid these problems. It's something more people should know!
Right? I thought this was common knowledge too; it’s crucial for managing cloud resources.
It's surprising how little this is discussed given how important health checks are!
Why not turn this into a GitHub issue instead of Medium? This seems like a legit problem that could benefit from a community fix.
That's the plan! But it appears maintainers prioritize other issues over this one.
I really think getting this on GitHub would help raise awareness and lead to a solution.
Thanks for your write-up! We're considering migrating to this setup, and it’s great to see firsthand accounts of issues. A GitHub issue might really help, as this seems crucial for many.
For sure! Even if it's specific, it's good to document these things. Someone might run into it later.
I did think about opening an issue, but I'm trying to gauge interest as it might just be an edge case.
It's great you managed to find a workaround! But posting your detailed solution on Medium isn't the best move if it's behind a paywall—many won’t see it. You might want to consider sharing it somewhere more accessible, like a GitHub issue or a community forum.
Yeah, I'm not a fan of Medium either; I typically avoid those 'member-only' articles.
Totally agree! If it's not easily accessible, it defeats the purpose.
That does make sense; I'll definitely look into implementing better monitoring.