https://github.com/osquery/osquery logo
#fleet
Title
# fleet
r

Robert Perrin

04/25/2023, 11:58 AM
Greetings! Shipping this thing to eks just fell into my lap. So far, I have struggled to just get it working with mysql and redis pods as the prereqs and migration job(s) seem to be kind of all over the place in the docs for k8s deployments. Is there a useful combination of the "Deploying to kubernetes" and the fleet helm chart? Or, do I just need to closely follow one deployment doc or the other? tl;dr what is the path of least resistance to k8s?
I know that is very vague. From what i've seen in the docs, my preferred architecture would look like fleet on eks, redis and sql native AWS, storing things in S3 for use in other tools and things.
perhaps contributing to the public helm chart may be the way 🙂
k

Kathy Satterlee

04/25/2023, 5:56 PM
Hey @Robert Perrin! Those guides could definitely use some love! I've reached out to the team for some input/best practices.
r

Robert Perrin

04/25/2023, 7:31 PM
Thanks, Kathy!! I am breaking the news that I have to reset expectations on this deployment hahaha...I want to do it right 🙂
b

Benjamin Edwards

04/26/2023, 3:11 PM
Hey Robert, I don't have a ton of K8s experience but I do have a lot of experience running Fleet in AWS. Personally I would stand up Elasticache (Redis) and Aurora RDS (MySQL) and then utilize K8s to simply handle the compute/application services. IMO the overhead of owning the dependencies is not something I'd want to take on. RDS & Elasticache have automated updates and backups, etc.
If K8s isn't a hard requirement, we actually have a terraform module that will stand up everything you need to run Fleet using ECS Fargate.
r

Robert Perrin

04/26/2023, 7:33 PM
Hey, Benjamin! Yeah, I saw that but it doesn't fit our patterns - SRE team runs the show, already have extensive RDS and EC presence, tf modules etc. So, i'll likely mvp with a downstream chart, and maybe pr the fleet chart if I do anything worthwhile 🙂
Thanks for the feedback and time
After struggling with eks deployments, i was able to move to use the serverless AWS pattern - with a little finagling of the /dogfood/ tf code, I was up and running fairly quickly with an mvp. Definitely don't recommend going eks. SSL offload on an ALB is so familiar and easy compared to learning how to get tls working on the tool itself.
infosec is happy to have their new toy. I will be working on solidifying my tf code, may bring PR to parameterize org differentiators and such for easy values.yaml style control
tfvars style ha