Big Bang 4.0: The Next Evolution of the Platform¶
Big Bang 4.0 is planned for October 2026. This release brings together sustained work across the Big Bang value stream to make secure Kubernetes platforms easier to operate, configure, and extend.
Many of these capabilities were introduced, evaluated, and refined throughout the Big Bang 3.x lifecycle. Big Bang 4.0 marks the point where that work comes together in four foundational areas: service mesh architecture, shared package integration, umbrella package configuration, and Kubernetes policy enforcement.
This article focuses on the value those capabilities provide. For detailed release and upgrade information, visit the Big Bang release notes and Big Bang documentation.
Istio Ambient Mesh Becomes the Default¶
Big Bang 4.0 makes Istio Ambient Mesh the default Istio configuration.
Ambient Mesh provides service-mesh security without requiring an Envoy sidecar
in every application pod. Instead, a shared, node-level proxy called ztunnel
provides Layer 4 connectivity, identity, and mutual TLS for workloads in the
mesh. When an application needs Layer 7 behavior, such as HTTP-aware policy or
routing, waypoint proxies can provide it selectively rather than imposing that
cost on every workload.
This architecture offers several practical benefits:
- Lower workload overhead: Proxy resource consumption no longer grows with every application pod in the mesh.
- Simpler operations: Application pods do not need to be restarted solely to receive routine proxy updates.
- Easier application onboarding: Mission applications can join the mesh without modifying pod specifications to inject a sidecar.
- Selective Layer 7 capabilities: Teams can apply deeper traffic processing where it provides value while using the lighter Layer 4 data plane elsewhere.
Supporting Ambient required coordinated work across the Big Bang platform.
Istio CNI, ztunnel, and Kubernetes Gateway API provide the underlying mesh
capabilities. Big Bang also updated its network and authorization policy model
to account for traffic carried over the HBONE tunnel. Together, these changes
preserve secure-by-default traffic controls as the data plane moves away from
sidecars.
Ambient first became available as an opt-in beta in Big Bang 3.23 and reached General Availability in Big Bang 3.32. It remains opt-in for the rest of the 3.x lifecycle before becoming the default in 4.0. We recommend that users begin migrating now with Big Bang 3.32 or later, starting in a development or test environment and validating mission-specific traffic and integrations before production. Migrating ahead of 4.0 lets teams evaluate the mesh change independently from the major-version platform upgrade.
Making Ambient the default does not remove sidecar mode: Big Bang 4.0 continues to support sidecar configuration so users can upgrade the platform without also being required to migrate their service mesh. If a mission environment uncovers an Ambient issue, sidecar mode provides a supported escape hatch while that issue is addressed. Maintaining both modes adds development and testing cost, so Big Bang may reconsider sidecar support in a future release through a separate deprecation decision. Read ADR 13 for the complete decision.
Authservice is also supported with Ambient through waypoint proxies, which
provide the Layer 7 enforcement that ztunnel cannot. When an inbound route
enables Authservice, bb-common creates a shared waypoint for the package
namespace along with the route’s JWT validation, external-authorization, deny
backstop, and required network policies. The consuming package enrolls the
protected Service onto that waypoint so both ingress and in-mesh traffic follow
the intended policy path. Big Bang’s Monitoring and Thanos integrations use
this route-scoped waypoint model for SSO-protected services.
Read the Ambient Mesh adoption overview to learn more about the architecture, or consult the Ambient migration guide for detailed preparation and compatibility information.
A Shared Integration Foundation with bb-common¶
As the Big Bang package ecosystem grew, packages often implemented the same platform concerns in different ways. Network policies were a clear example: similar rules could have different configuration shapes, defaults, and override behavior depending on the package. That duplication made package integration harder to understand, test, and maintain.
bb-common provides a shared Big Bang integration layer for these cross-cutting
concerns. The bb-common implementation focused on network policy, reusable
definitions, consistent default-deny behavior, and a concise domain-specific
language for expressing communication between workloads. Policy configuration
remains visible in package values, making the intended security boundaries
easier to review and audit.
This shared behavior is especially important for Ambient Mesh. Workloads using
Ambient communicate through the HBONE tunnel, and their network and
authorization policies must account for that traffic without weakening
segmentation. Centralizing those patterns in bb-common allows Big Bang to
apply them consistently instead of solving the same problem independently in
every package.
More broadly, a shared integration layer means that improvements and security
fixes can be implemented once and adopted across the package ecosystem. It
reduces duplicated templates, makes package behavior more predictable, and
provides one place to build comprehensive tests for common platform behavior.
Read Streamlining Integration with bb-common
for a deeper look at the design and network-policy model.
From Helm Library Chart to Regular Subchart¶
Big Bang is also standardizing how packages consume bb-common. Packages
initially used it as a Helm library chart, selectively calling shared templates
through package-specific include helpers. In that model, Big Bang-specific
values such as istio, networkPolicies, and routes were supplied at the top
level of the package’s values.
Packages are moving to consume bb-common as a regular Helm subchart. This
places its configuration under a scoped bb-common key and uses Helm’s standard
dependency behavior. The Big Bang umbrella currently supports both shapes
while integrated packages complete the transition.
The move makes package integration simpler, safer, and more consistent. As a
regular subchart, bb-common can enforce a strict values schema and catch
incorrectly structured configuration before deployment. This reduces the risk
of accepting configuration that is valid YAML but silently has no effect. It
also removes the need for packages to add integration-specific include
templates: package authors can add bb-common as a standard chart dependency
and supply values through its scoped configuration.
The result is less custom integration code, clearer ownership of Big Bang-specific behavior, and a common dependency model across the package ecosystem. It also brings integrated packages into alignment with the subchart model already used by team-maintained packages. The complete architectural decision is recorded in ADR 14.
During the transition, the umbrella chart merges common defaults with user
overlays and supplies the correct values shape for each package. Packages that
have migrated receive a nested bb-common configuration, while packages still
using the library chart continue to receive the legacy flat values. Every Big
Bang integrated and team-maintained package will complete this migration for
4.0, allowing the umbrella to remove its temporary compatibility logic and use
the standard subchart model consistently.
This transition changes the values shape for these packages, so existing users
will need to migrate their package overrides. For the 4.0 upgrade, users will be
able to use Big Bang’s
3-to-4 values migration script for both
breaking configuration changes. In one migration, the script will move legacy
top-level and addons.<name> package configuration into the consistent
packages.<name> map and rewrite integrated and team-maintained package
overrides from the Helm library-chart values shape to the scoped bb-common
subchart shape. Users should review and validate the generated values before
deployment.
Bring your own packages and mission applications remain under their owners’
control: packages already using bb-common as a subchart do not need this
consumption-model migration, while owners using the library pattern remain
responsible for updating their package and values. Big Bang recommends subchart
consumption when an owner can modify or wrap the application chart. The
bb-common integration guide
also documents Argo CD and Kustomize alternatives for applications where adding
the subchart is not possible.
One Consistent Way to Configure Every Package¶
Big Bang has historically exposed package configuration through three different
locations. Core packages use top-level keys such as monitoring, add-on
packages use keys such as addons.gitlab, and user-supplied packages use
packages.<name>. As a result, users and automation must know how Big Bang
categorizes a package before they can locate its configuration.
Big Bang 4.0 introduces one canonical model: built-in and user-supplied packages
are configured under packages.<name>.
For example, package enablement moves from separate core and add-on locations:
# Big Bang 3.x legacy paths
monitoring:
enabled: true
addons:
gitlab:
enabled: true
packages:
confluence:
enabled: true
bringYourOwnPackage:
enabled: true
to one consistent package map:
# Big Bang 4.x unified package configuration
packageConfiguration:
version: v1
packages:
monitoring:
enabled: true
gitlab:
enabled: true
confluence:
enabled: true
bringYourOwnPackage:
enabled: true
This provides one predictable location for every deployable package. Automation no longer needs special logic for core, add-on, and user-supplied packages, and a package can change category without moving its public configuration path. Global platform configuration—including domain, registry credentials, shared network-policy settings, and shared Istio configuration—remains at the top level.
The unified model is backed by a repository-owned package metadata catalog. That catalog defines canonical package identities and generates the relevant schema and migration mappings. It allows Big Bang to detect ambiguous package names and resource-name collisions earlier while reducing duplicated package metadata across the umbrella chart and its tooling.
Users can adopt the unified contract during the 3.x lifecycle with
packageConfiguration.version: v1. In Big Bang 4.0, v1 becomes the default
and supported package contract, and the legacy top-level and addons.<name>
package paths are removed. Read
ADR 11
for the design decision and the
package values migration guide
when preparing existing configuration.
Kyverno Policies Move to CEL¶
Big Bang 4.0 modernizes its policy foundation by moving the
kyverno-policies package away from Kyverno’s legacy ClusterPolicy resources
and onto its purpose-built, CEL-based policy APIs.
Big Bang currently integrates Kyverno v1.19, which still supports
ClusterPolicy. Upstream Kyverno identifies v1.19 as the final release with
full support for its legacy policy types: ClusterPolicy and Policy are
officially deprecated in v1.19 and will be removed in v1.20. Kyverno currently
estimates the v1.20 release for November 2026, and Big Bang expects to integrate
Kyverno v1.20 through the Kyverno package in November. Moving Big Bang policies
to CEL as part of 4.0 prepares users for that package update before the legacy
resources disappear. See Kyverno’s
v1.19 release announcement
and policy-type deprecation schedule
for the upstream timeline.
Common Expression Language (CEL) is used throughout Kubernetes for expressions and admission control. By adopting Kyverno’s CEL policy types, Big Bang aligns its policy implementation with that broader Kubernetes direction and moves from one resource that mixes multiple rule types to APIs with clearer responsibilities:
ValidatingPolicyallows or rejects resources during admission.MutatingPolicymodifies resources.GeneratingPolicycreates related resources.ImageValidatingPolicyverifies container images.
Big Bang’s built-in policies retain recognizable names and package-level configuration even though the underlying resource types and expression model change. This preserves a familiar policy catalog while moving its implementation onto a more modern foundation.
Policy exceptions evolve with the policy resources. Legacy exceptions referred
to policy and rule names and selected resources through the older match model.
The CEL-based PolicyException uses policyRefs to identify policies and CEL
matchConditions to select resources. Big Bang provides package-specific
exceptions for workloads that legitimately require elevated permissions or
access to the Kubernetes API.
The new configuration schema rejects known legacy fields so incompatible overrides fail visibly instead of being silently ignored. Users who deploy the default policy configuration receive the updated policy foundation through Big Bang. Users who customize policies, exclusions, or exceptions should review the upstream CEL migration guide before adopting 4.0 and follow the Big Bang release notes for package-specific guidance.
Preparing for Big Bang 4.0¶
These capabilities change important platform defaults and configuration contracts. Existing users should review the Big Bang 4.0 release notes and the linked migration guides before upgrading. Detailed compatibility requirements, breaking changes, and step-by-step actions will be maintained in those sources rather than repeated here.
Built with the Community¶
Big Bang 4.0 replaces bespoke and legacy patterns with platform foundations
that are simpler, more consistent, and aligned with Kubernetes and Helm
standards. Ambient Mesh reduces the operational weight of the service mesh,
bb-common makes secure package integration reusable, unified package
configuration gives users one coherent contract, and CEL provides a modern
policy foundation.
Together, these capabilities reduce platform friction so teams can spend more time securely delivering mission applications.
Thank you to the package maintainers, contributors, testers, and community members whose feedback shaped this work. Follow the Big Bang release notes and Big Bang documentation for the latest information as the October 2026 release approaches.