Hi all. Got asked today from one of our directors. He asked me few questions • "How can we see the health of the nodes in a cluster after it's deployed by Pulumi? Currently, we can easily see the health status of our nodes in Spinnaker" • "How can we redeploy back or rollback to the previously deployed resources?" In Spinnaker, we can either "Restart pipeline of the previous known-good build" or "rollback to the sleeping ec2 instance, redblack deployment". I told him I'll do some research. What are you using for the monitoring of your nodes? How did you design your rollback strategy? Did you use any third party tool(s)?

“How can we see the health of the nodes in a cluster after it’s deployed by Pulumi? Currently, we can easily see the health status of our nodes in Spinnaker”

“How can we redeploy back or rollback to the previously deployed resources?” In Spinnaker, we can either “Restart pipeline of the previous known-good build” or “rollback to the sleeping ec2 instance, redblack deployment”.

Pulumi isn’t a replacement for Spinnaker, it would work alongside spinnaker. If you’re moving away from Spinnaker, you’d likely want to see the health of the nodes in whatever replacement you use. Same applies for rollbacks. Spinnaker is a contiuous deployment tool, Pulumi is an Infrastructure as Code tool

Yep, true. I would like to know what you folks are using for cd. Spinnaker has been a headache for us for the past 2 years now. So we're going to use gitlab-ci for the ci and cd. However, for the monitoring, we don't know yet. I would like to hear what others are using.

there’s a hole bunch of options here. Really depends on your needs and requirements. Which cloud are you using?

AWS. They can monitor it using AWS but we want to give them as little access as possible. Maybe, we just need to limit their access to AWS.

Code Deploy is AWS’s continuous delivery tool. I’ve also enjoyed using harness.io before. What does your compute run on? ECS? EC2 instanaces? EKS?

Most right now are on EC2 and very few on ECS and much more fewer on EKS. Dev teams wants to move their apps to containers so we are working with AWS to help our devs convert their apps as containers

I see that some people have got it working, for example: https://medium.com/datapeaceai/configuring-gitlab-ci-cd-with-aws-codedeploy-and-docker-for-python-application-8ae0e9912771

wow, cool! Our aws support is wrong, lol 😄 Thank you

Well, not wrong.. you essentially have to implement everything that the GItHub action does yourself, and you miss out on AWS support (for that bit of the pipeline) because of it. So it "works" on GitHub, and you "can get it working" on GitLab 🙂

got it

Personally, I found a combination of a pipeline service (Bitbucket Pipelines, in my case) + FluxCD to create a GitOps deployment process, and suitable Monitoring & Alerting via Grafana, etc. to be far better than Spinnaker. What I think spinnaker did make easy was interdependent jobs, especially where you want a prerequisite job like a Db migration (E.g. Flyway) to run before the main deployment but, that can now be solved in various ways and so the benefit is negligible. ArgoCD might also give you a bit more of Spinnaker-like visibility/dashboarding in a GitOps style deployment, if that's what you need.

Thank you for sharing your setup. I was told about ArgoCD before that it can only be used if we have our own kubernetes(not AWS EKS). That's why I stopped playing with it. I remember seeing a demo of it in 2018 I think at KubeCon in San Diego! I really WOWed!

Ah ok, I didn't know about that limitation. Do you happen to know what the restriction was?

That’s not a limitation, you can use ArgoCD with eks

Oh wow, the person who told me last year about it does not know it then. That's great to know!

That's what I thought but I didn't want to contradict as I didn't know for sure. I couldn't think of a reason why Argo would need direct access to nodes or the control plane so couldn't think of a reason why it wouldn't work.

Cool! Then I'll start playing with ArgoCD again 😄

🙂