Gabriel Coutu
FrançaisContact

CONSULTANT · OBSERVABILITY · SRE · PLATFORMS

Clearer production. More reliable platforms.

I help technology teams improve observability, diagnose production issues, and evolve Kubernetes platforms—with choices suited to their context, from cloud to bare metal.

Discuss a projectExplore my experience

Based in Montreal · Quebec and remote

EXPERIENCE AT SCALE

01~200servers in production environments
0210+clusters managed with Cluster API
03~4 TBof logs over three weeks
04~600 GBof metrics over two months

HOW I CAN HELP

From diagnosis to knowledge transfer.

Focused engagements, explained decisions, and deliverables shaped for your team.

/01

Observability and reliability audit

Map metrics, logs, and alerts; identify blind spots; prioritize improvements against your risks and constraints.

/02

Monitoring, logs, and alerting

Design or evolve telemetry pipelines, dashboards, and alerts—including improving the solutions you already use.

/03

Kubernetes and GitOps architecture

Build reproducible platforms, multi-cluster management, and declarative deployments across cloud, hybrid, or bare-metal infrastructure.

/04

Incident diagnosis and SRE support

Add focused capacity to investigate a production issue, improve operational practices, and share methods with your team.

EXPERIENCE IN CONTEXT

Real production challenges, at meaningful scale.

Experience is anonymized. Migrations described here were progressive.

CASE / 01OBSERVABILITY

Modernizing observability across a production fleet

Evolve host-by-host monitoring into centralized metrics and log pipelines suited to significant data volumes.

CONTEXT

An organization operating approximately 200 servers relied on Nagios and ad hoc SSH access. Metrics and logs came from separate systems, complicating investigation and monitoring improvements.

WORK

Contributed to a progressive evolution toward Prometheus and Thanos, then VictoriaMetrics for metrics. For logs, Vector was used as Loki evolved toward Quickwit. Grafana and Alertmanager supported exploration and alert handling.

  • Approximately 600 GB of metrics retained over a two-month window.
  • Approximately 4 TB of logs processed over a three-week window.
  • Observability pipelines and tooling modernized progressively.
ARCHITECTURE / TRANSITION

BEFORE

  1. Nagios
  2. Ad hoc SSH
  3. Fragmented monitoring

EVOLUTION

  1. 01Hosts and services
  2. 02Centralized collection
  3. 03VictoriaMetrics + Quickwit
  4. 04Grafana + alerts
TECHNOLOGIES
PrometheusThanosVictoriaMetricsVectorLokiQuickwitGrafanaAlertmanager

Volumes are indicative and reflect the reported retention windows. This case study is anonymized and does not claim that one engagement alone produced every operational outcome.

CASE / 02PLATFORM ENGINEERING

Progressively evolving a legacy platform to Kubernetes and GitOps

Gradually replace a custom deployment stack with a more declarative and reproducible Kubernetes platform.

CONTEXT

A production platform relied on LXC containers, custom Debian packages, and Perl scripts. Evolving and reproducing it required substantial manual effort.

WORK

Contributed to a progressive migration to bare-metal Kubernetes with FluxCD, Helm, and GitOps practices. SOPS for secrets, Velero backups, and Ceph storage were part of the ecosystem. Also experienced a fleet of more than ten clusters managed with Cluster API.

  • Transition proceeded in stages, with legacy and Kubernetes platforms coexisting during the evolution.
  • Experience with a Kubernetes environment of approximately 20 nodes.
  • Management of a 10+ cluster fleet with Cluster API, described separately from the approximately 20-node environment.
ARCHITECTURE / TRANSITION

BEFORE

  1. LXC containers
  2. Custom Debian packages
  3. Perl scripts

EVOLUTION

  1. 01Declarative Git
  2. 02FluxCD + Helm
  3. 03Kubernetes
  4. 04Ceph + Velero
TECHNOLOGIES
KubernetesFluxCDArgo CDHelmCluster APISOPSVeleroCeph

The scope evolved in stages. The approximately 20 nodes and fleet of more than ten clusters describe distinct experience and are not combined.

TOOLS AND PRACTICES

Systems thinking, not a buzzword list.

DOMAIN / 01

Observability and metrics

Collection, storage, visualization, and alert management.

  • Prometheus
  • VictoriaMetrics
  • Thanos
  • Grafana
  • Alertmanager
  • Grafana OnCall
  • incident.io
DOMAIN / 02

Logs and telemetry pipelines

Collection, transformation, routing, and log search.

  • Vector
  • Quickwit
  • Loki
  • OpenTelemetry
DOMAIN / 03

Platform, GitOps, and data

Orchestration, cluster lifecycle, deployments, and storage.

  • Kubernetes
  • Cluster API
  • FluxCD
  • Argo CD
  • Helm
  • Cilium
  • SOPS
  • Velero
  • Ceph
  • PostgreSQL
  • ClickHouse
DOMAIN / 04

Systems and engineering

Linux foundations and operational problem solving.

  • Linux
  • Debian
  • Bare metal
  • Automation
  • Performance diagnosis
  • Incident management

HOW I WORK

Pragmatic choices, with a clear handoff.

01

No black box

Versioned configuration, useful documentation, and knowledge transfer. The aim is to help your team evolve the solution.

02

The right trade-off for your context

Managed service, existing tool, or dedicated component: choices based on needs, team skills, total cost, and operational effort.

03

More signal, less noise

Alerts tied to clear action and telemetry that helps diagnosis—instead of more dashboards without decisions.

START A CONVERSATION

A platform or reliability challenge to untangle?

Send a few lines about your context. We can see whether my experience fits your needs.

Send an emailMontreal, Quebec