Self-hosted observability · London, UK
I help teams run their own metrics and alerting without it becoming the slowest, noisiest or most expensive thing they own.
I'm Korhan Ozturk, a data and platform engineer. Koz Software is my consultancy. I specialise in observability for shared infrastructure: Prometheus and Grafana Mimir platforms that serve many teams at once.
What I work on
Shared observability platforms
Designing, running and fixing self-hosted metrics platforms that many teams share: Prometheus, Grafana Mimir and Grafana. That covers multi-tenancy, limits, sizing and upgrades.
Metrics and alerting that hold up
Cardinality under control, recording rules that do not overload the ruler, and alerts that point at real problems. Fewer, better signals that people act on.
Cost and capacity attribution
Finding out which tenants, teams and rules are using your metrics infrastructure, then turning that into showback, sensible limits and a capacity plan.
Data and platform engineering
The pipelines and platforms underneath: data movement, Kubernetes, infrastructure as code, and the build and deployment tooling that keeps it all running.
Currently building
obscostIn active development
Cost allocation for self-hosted Grafana Mimir. It tells platform teams which tenants and which rules are using their shared metrics infrastructure, from what the cluster actually measured rather than estimates.
$ promcost cost \
--metrics-tenant monitoring \
--tenant analytics,infra,payments,platform \
--since 7d
analytics uses 62.6% of ingester memory
the equivalent of 5.0 of your 8 ingesters.
tenant share ≈ ingesters
analytics 62.6% 5.01
infra 32.3% 2.59
payments 5.0% 0.40
platform not measuredAbout
Most of my work has been on the platform side of observability: the shared metrics and alerting systems a central team runs for everyone else. It's a job where the technical problems (cardinality, query load, retention, scaling) come tangled up with organisational ones: who owns this alert, who is paying for those series, which team's rules are slowing everything down.
I work directly with engineering teams, hands-on, from a single session to a longer engagement. More about me