AI-generated editorial illustration; not a documentary photograph.
GitLab’s September 22, 2026 guide outlines how organizations should plan enterprise deployments around infrastructure control, CI/CD workload demand, recovery objectives and performance testing. It presents architecture decisions rather than a new product release, with the main tradeoff being how much operational responsibility a team can sustain.
Three operating models
The guide distinguishes GitLab.com, GitLab Dedicated and GitLab Self-Managed. GitLab.com is a multi-tenant SaaS service, while GitLab Dedicated is a fully managed single-tenant offering hosted on Amazon Web Services. Self-Managed gives the customer control of installation and infrastructure, but also responsibility for scaling, security, upgrades, monitoring and recovery.
GitLab recommends documenting requirements such as data residency, network isolation, recovery objectives and infrastructure control before selecting a model. The choice should also account for which team will own backups, capacity planning and incident response.
Size runners by workload
A central recommendation is to size GitLab Runner capacity from CI/CD behavior rather than developer headcount. Job volume, duration, peak concurrency, compute requirements, network paths and sensitive workloads determine how many jobs must run concurrently and how long queued work can wait.
Runner scope affects both isolation and utilization. Instance runners serve projects across an installation, group runners cover a defined group, and project runners support dedicated credentials or specialized infrastructure. GitLab-hosted runners can shift provisioning, patching and autoscaling work away from customer teams on GitLab.com and GitLab Dedicated.
Recovery needs separate layers
The guide treats high availability, disaster recovery and backups as different protections. Self-Managed operators must turn recovery targets into architecture requirements and rehearse failover, including dependencies such as identity and networking. GitLab Geo provides active-passive disaster recovery for Self-Managed environments, but customer-managed operational steps remain part of failover.
For GitLab Dedicated, GitLab manages the underlying disaster recovery infrastructure and failover process. Customers can select a secondary AWS region for geo-based recovery, with replication between the primary and secondary regions managed by GitLab.
Measure bottlenecks before scaling
GitLab separates queued duration, job execution time and overall pipeline duration to help locate constraints. Queue delays may point to runner capacity, while slow execution can reflect pipeline design, test suites, dependency downloads or repository transfers. The guide recommends testing representative projects during normal and peak conditions, including release windows and heavy commit activity.
Kubernetes introduces another operating decision: its executor runs CI/CD jobs as pods in an existing cluster, while GitLab Self-Managed can run on Kubernetes as part of a cloud-native architecture. These roles have different capacity, observability and recovery implications. GitLab’s Cloud Native reference architecture keeps PostgreSQL, Redis and object storage external, while teams needing Gitaly Cluster should evaluate hybrid or VM-based designs.
Source: How to design GitLab for enterprise scale
SIGNAL TITANS
Leave a Reply