Infrastructure through Prometheus
No new exporter, password or network path is needed: Varqeni reads what Prometheus already scrapes.
AvailabilityAvailable in early access. Infrastructure Operations changes nothing on its own; its four fixes run only after a named approval.
How it works
- What it reads: the hypervisor and its guests and storage, node exporter on each host, Kubernetes state, object storage, the database exporter, Kafka brokers and topics, Kubernetes autoscalers and load balancers, and certificate probes.
- What it does: names one cause for related problems, such as a stopped virtual machine and its node going not ready, and recommends what to do. Four fixes run after a named approval: restart a host's metrics exporter, clear old logs from a full disk, restart a hung workload, roll back a bad rollout.
- Kafka: a broker that is down or fenced out of the cluster, no active controller, offline partitions, partitions below their minimum in-sync replicas, under-replicated partitions, and a topic you list that stops receiving messages for five minutes. A broker outage is one cause, with the partitions it leaves short listed as its consequences; topics kept on a single replica are shown as a watch item. Read-only: Varqeni never restarts a broker. In the lab it named a fenced broker or a stalled topic correctly in 30 of 30 blind lab trials.
- Autoscaling and load balancers: an autoscaler held at its maximum while it still wants more, an autoscaler that cannot read the metric it scales on, a load balancer with no external address, and one with no ready backends. A crash loop behind a load balancer stays the cause, with the empty load balancer as its consequence. Read-only: raising a maximum is a capacity decision for a person. 30 of 30 blind lab trials.
- What it cannot do: act on its own, or start a stopped virtual machine (built and switched off).
What you should see: The estate under Infrastructure in the console.
What it can and cannot do
Can
- Read the metrics your exporters already collect, Kafka brokers and topics, autoscalers and load balancers included, and name one cause for related problems.
Cannot
- Change infrastructure without a named approval, restart a Kafka broker, raise an autoscaler's maximum, or watch a host that no exporter reports on.
Related articles
- Read the estate under InfrastructureUnder Infrastructure, the console shows hosts, Kubernetes, data services, Kafka, autoscalers, load balancers, storage pools and certificates, with any findings.
- Data, Infrastructure and Service areasThe area buttons next to the views switch between Data Operations, Infrastructure Operations and Service Operations.
Still stuck? Contact support or email info@fortisdatagroup.com.