On this page
Quick answer
RubixKube watches every place your systems run with read-only observers: Kubernetes clusters, AWS, Google Cloud and Azure, and Linux machines. Alerts from Datadog, PagerDuty, Grafana, Sentry and other tools come in by webhook. Rubix investigates across all of it, so one incident can follow a failure from a cloud database to the pod it broke.
Production is rarely one cluster. It is a few clusters, a cloud account or two, some machines nobody wants to touch, and five monitoring tools. An AI SRE that sees only one of them sees only part of every incident.
What you can connect
Sources
| Source | How it connects | What it watches |
|---|---|---|
| Kubernetes | One kubectl apply from Environments > Connect. EKS, GKE, AKS, self-managed and KIND | Pods, workloads, Services, Ingresses, nodes, events |
| AWS | One script, on your host or on a new small EC2 instance it creates | EC2, EBS, VPC, RDS, S3, Lambda, ELB, EKS, IAM, CloudWatch, CloudTrail |
| Google Cloud | The same script, on your host or a new small instance | Compute Engine, GKE, Cloud SQL, Cloud Run, Functions, Storage, Pub/Sub, Monitoring, Logging |
| Azure | The same script, or AKS as a Kubernetes cluster | AKS clusters, virtual machines and the services that run on them |
| Linux machines | A small service on Linux with systemd, or macOS | CPU, memory, disk, swap, load and stuck processes, every 30 seconds |
| Alerting tools | Webhook: Datadog, PagerDuty, Grafana, Sentry, Jira, Linear, Observe or your own | Their alerts become incidents |
Rubix can also read your other tools while it investigates, such as New Relic, Honeycomb, Cloudflare, GitHub, GitLab, Notion and Linear, through the integrations catalog.
One picture, not five
- Sources and environments. Each cluster, account, machine or tool is a source. Label sources with an environment such as production or staging, and filter everything by it.
- Linked sources. When two sources describe the same system, such as Datadog monitoring the cluster your observer watches, RubixKube links them. Shared ids confirm a link in code. Name matches come to you as a Yes or No question on the source's page.
- One incident across tools. Linked sources' alerts land in the same group, so the Datadog alert and the observer's finding meet on one incident with one RCA.
Read-only, and light
- Observers read. The cloud observers use read-only and viewer roles. The Kubernetes observer reads ConfigMaps and Secrets as metadata only.
- Small footprint. The Kubernetes observer is one pod, requesting 256Mi of memory and 250m of CPU, sized for about 50 nodes.
- Personal data is masked before it reaches the model, and your data is kept in Google Cloud's us-central1 region.
Plans
The free Individual plan covers one environment. Business covers three, and Enterprise is set for your organization. See pricing.
Start free, or read Connect your environment, Kubernetes, AWS, Google Cloud and machines.
Frequently asked questions
Does RubixKube replace Datadog or Grafana?
No. It brings its own observers, and works with the tools you already have: their alerts become incidents and Rubix reads their data while it investigates.
Does it support Azure?
Yes. Connect AKS clusters with the Kubernetes install, and Azure virtual machines with the same script as the other clouds.
Which machines does the observer run on?
Linux with systemd and macOS, on x86_64 and arm64, as a small background service.
How long does setup take?
A Kubernetes observer usually connects within a minute. Cloud observers connect in a minute or two.