Project Name
From Manual Deployments to GitOps: How a SaaS Platform Cut Release Time by 60% With ArgoCD
A Series B SaaS company building a B2B analytics platform for enterprise financial services and healthcare customers ran 12 microservices on Kubernetes across three environments, deployed by senior engineers running Helm and kubectl commands over SSH after receiving a ticket. Every release consumed 4-5 hours of coordinated senior engineering time, three of the previous quarter’s five production incidents were traced directly to deployment errors, and the VP of Engineering couldn’t answer “what’s deployed where right now” without interrupting someone.
Ksolves was trusted as a Kubernetes development partner to assess the existing release process and design a GitOps architecture around ArgoCD with environment-specific ApplicationSets, training the internal platform team from assessment to production in eight weeks.
- Manual, Ticket-Driven Deployments Taking 4-5 Hours Per Release: Every release required a senior engineer to SSH into the cluster, run Helm upgrade commands, verify pod health, and update the ticket by hand, serializing every deployment across all 12 microservices.
- Frequent Human Errors Causing Production Incidents: Manual deployments introduced a steady stream of errors, wrong image tags, overwritten environment values, incorrect namespace targets, and three of the previous quarter's five production incidents traced directly to deployment mistakes.
- No Deployment Visibility for Engineering Leadership: The VP of Engineering had no real-time view of what was deployed where or whether cluster state matched Git, and enterprise customer audit questions required manual investigation every time.
- Configuration Drift Between Environments Going Undetected: With no reconciliation layer, cluster state in staging and production regularly drifted from what Git declared, as engineers applied hotfixes directly to pods with no corresponding commit.
- Senior Engineer Bottleneck Blocking Release Velocity: Only three engineers had the cluster credentials and Helm knowledge to execute production deployments, gating every release on one of them being available.
- No Deployment Audit Trail for Enterprise Compliance: Enterprise financial services and healthcare customers were beginning to request deployment audit logs as part of vendor security reviews, and the team had no way to produce one.
Ksolves ran a two-week discovery assessment of the existing release process before proposing an architecture centered on ArgoCD as the deployment controller, with GitHub Actions CI left unchanged and a separate manifests repository as the single source of truth for cluster state.
- Discovery and Release Process Assessment (Weeks 1-2): The existing release process was mapped end-to-end, every manual step documented, every error mode catalogued, and the root cause of the three deployment-related incidents analyzed into a prioritized list of failure modes.
- Manifests Repository Design and Helm Chart Restructuring (Weeks 3-4): A dedicated manifests repository, separate from application source code, was created with Helm values files per environment for all 12 microservices, and existing charts were restructured to eliminate hardcoded environment-specific values.
- ArgoCD Installation and ApplicationSet Configuration (Weeks 4-5): ArgoCD was deployed with OIDC SSO integration, Casbin-based RBAC, and an ApplicationSet list generator creating one Application per microservice per environment, with auto-sync for dev and staging and manual-sync-only for production.
- GitHub Actions CI Pipeline Integration (Weeks 5-6): The existing CI pipeline was extended with a final step opening a PR to the manifests repository with the new image tag, triggering ArgoCD's sync of the staging Application, with production promotion requiring explicit lead approval.
- Drift Remediation and Cluster Cleanup (Week 6): Before enabling self-heal, every difference between live cluster state and declared Git state was audited across all environments, resolving 34 configuration drift items as reviewed PRs.
- Platform Team Training and Documentation (Weeks 7-8): Four hands-on workshops trained the internal platform team on ArgoCD operations, RBAC administration, and incident response under the GitOps model, backed by a runbook and adoption guide.
Technology Stack
| Category | Technology |
|---|---|
| GitOps CD | ArgoCD + ApplicationSet |
| CI Pipeline | GitHub Actions |
| Source Control | GitHub (App Repo + Manifests Repo) |
| Packaging | Helm |
| Authentication | GitHub OIDC + ArgoCD RBAC |
| Observability | ArgoCD UI + Notifications Controller |
- Release Time Cut 60%, From 4-5 Hours to Under 90 Minutes: The GitOps flow from merged PR to staged deployment now completes automatically in under 15 minutes, with total time from build to production under 90 minutes.
- Zero Deployment-Related Incidents Post-Go-Live: ArgoCD's declarative sync eliminated the entire class of errors caused by manual kubectl and Helm commands, with zero deployment-related production incidents in the three months following go-live.
- Release Frequency Up From Twice a Week to Daily: With automated GitOps deployment and environment-specific sync policies, individual microservices now deploy independently and daily, cutting batch size and blast radius for any given feature.
- Real-Time Deployment Visibility for Leadership: ArgoCD's web UI now provides real-time sync status, health, and deployment history for every one of the 12 services across all three environments, accessible to any SSO-authenticated stakeholder.
- 34 Drift Items Remediated, Audit Trail Established: ArgoCD self-heal now reconciles cluster state with Git continuously, catching drift within minutes, with every cluster change recorded as a reviewed Git commit.
- Senior Engineer Bottleneck Removed: Any engineer with team-lead RBAC can now trigger a production sync by merging a reviewed PR, replacing a bottleneck on execution capability with an appropriate approval gate instead.
- Compliance Audit Trail Delivered: ArgoCD's sync history now satisfies the deployment audit requirements enterprise financial services and healthcare customers had started requesting in vendor security reviews.
“We used to spend half a sprint just moving code from staging to production. Now that happens automatically while the team is building the next feature; ArgoCD gave us back the engineering capacity we needed actually to hit our roadmap commitments.”
– VP of Engineering, SaaS Platform
Manual, ticket-driven Kubernetes deployments on this platform took 4-5 hours per release, caused three of the previous quarter’s five production incidents, and left engineering leadership with no real-time view of what was actually running anywhere. Kubernetes development services from Ksolves built a GitOps architecture on ArgoCD with a two-repository manifest model and environment-specific ApplicationSets, delivered from assessment to production in eight weeks.
Release time dropped 60% to under 90 minutes, deployment-related incidents fell to zero, and release frequency rose from twice a week to daily. The senior engineer bottleneck that gated every release is gone, replaced by an appropriate approval process instead of a scarce execution capability, and the company now has the compliance audit trail its enterprise customers had started asking for.
This GitOps foundation scales linearly with the engineering organization; adding a new microservice means adding a values file and an ApplicationSet entry, with no changes to process or tooling, and positions the team for progressive delivery with canary deployments as the next step.
Is Your Engineering Team Still Spending Hours on Manual Deployments With No Real Visibility Into What’s Running in Production?
