# Tekton v1.13.0

2026-05-29 · [Original release notes](https://github.com/tektoncd/pipeline/releases/tag/v1.13.0) · [Full summary](https://ratatosk.io/en/releases/tekton/v1.13.0)

Security 3 · Breaking changes 2 · Important fixes 6 · Enhancements 1 · Dependency updates 32 · Other changes 17

## Security (3)

### [Security status unclear] TaskRun notification tracing added

- Affects: TaskRun notification reconciliation and CloudEvent delivery paths, including the shared `ReconcileRunObject` path used by `CustomRun`
- Additional action: Trusted observability systems for exported traces, which may include Kubernetes resource identifiers.

Previously, failures could not be correlated or latency measured between the reconciliation trigger and final CloudEvent emission. Linked spans now trace this path; the TaskRun root span includes `taskrun.name` and `taskrun.namespace`, and the CloudEvent span records network and sink failures.

> * :sparkles: feat(tracing): add spans to TaskRun notifications controller (#9912)
> Added OpenTelemetry tracing to the TaskRun notification reconciliation path.
> Spans now cover ReconcileKind, ReconcileRunObject, and EmitCloudEvents,
> enabling operators to trace CloudEvent delivery latency end-to-end.

### `TaskRun` notification identifier logs moved to debug level

- Affects: `TaskRun` notification reconciliation
- Additional action: For exported traces, a trusted backend with access controls equivalent to or stricter than the CloudEvents sink; in multi-tenant environments, reviewed retention and access-control policies before enabling tracing.

Previously, this path logged `TaskRun` and run-object identifiers at INFO by default. After upgrading, these identifiers are no longer emitted at the default log level, without operator changes.

> * :bug: fix(events): reduce TaskRun notification identifier logging and document trace exposure (#10117)
> Reduced default log verbosity in the TaskRun notification path introduced
> by #9912. Identifier logs (TaskRun name/namespace) are now emitted at debug

### Resolver validation now rejects unsupported objects

- Affects: Custom resolvers or ResolutionRequests using the Resolver API

Resolvers could previously fetch any object or string and mark the ResolutionRequest successful; validation happened only when it was injected into the parent Run. Now, resolving anything other than a Tekton StepAction, Task, or Pipeline fails the ResolutionRequest before the data is written to it.

> * :bug: fix(resolvers): validate data is Tekton object in resolver framework (#9588)
> Fixes a bug which lets Tekton Resolvers resolve non-tekton objects and arbitrary data. After this change, resolving a non-tekton object causes the ResolutionRequest to fail.
> Action Required: Tekton Resolvers are now only permitted to resolve StepActions, Tasks, and Pipelines. Custom resolvers or ResolutionRequest which use the Resolver API for other object types will no longer function.

## Breaking changes (2)

### Configured `config-events` formats now control event emission

- Affects: setups with the `formats` field configured in `config-events`

With no configuration change, the `tektonv1` default preserves existing behavior. An invalid or unrecognized format logs a warning and suppresses event emission for that format.

> * :sparkles: [TEP-0137] Activate formats field in config-events (#9776)
> The formats field in config-events is now active. The default value is
> tektonv1, which preserves existing behaviour. Setting an invalid or
> unrecognised format value logs a warning and suppresses event emission
> for that format.

### Anonymous Pipeline labels and metrics use sanitized `generateName`

- Affects: inline-spec Pipelines with `metadata.generateName` set

Previously, TaskRuns from these runs used the PipelineRun name for the `tekton.dev/pipeline` label and the `pipeline` metrics tag.

> * :bug: fix(pipelinerun): use generateName for anonymous pipeline label (#9826)
> Previously, TaskRuns' `tekton.dev/pipeline` label for anonymous Pipelines would use their PipelineRun's name in their `tekton.dev/pipeline` label and `pipeline` metrics tag.  After this change, TaskRun and PipelineRun which are created from anonymous Pipelines now reference a sanitized `metadata.generateName`, when present, for the label `tekton.dev/pipeline`. Similarly, these TaskRruns and PipelineRuns will populate their respective metrics' `pipeline` tag with the sanitized generateName instead of using `anonymous`.

## Important fixes (6)

### Previous TaskRun condition context preserved on PipelineRun timeout

- Affects: TaskRuns managed by PipelineRuns with timeouts

Previously, a PipelineRun timeout could overwrite a child TaskRun's `status.conditions`, losing details such as `ExceededResourceQuota`. The new failure message includes the previous condition's reason and message unless its reason is `Started`, `Running`, or `Pending`.

> * :bug: fix: preserve previous condition context when TaskRun is cancelled or times out (#10076)
> Preserve previous TaskRun condition context (reason + message) when a TaskRun is cancelled or times out due to a PipelineRun timeout, so diagnostic information like ExceededResourceQuota is no longer lost.

### Resolver cache now honors resolver-specific TTLs

- Affects: resolver ConfigMaps with a resolver-specific TTL greater than 0, such as `bundleresolver-config` or `git-resolver-config`

Previously, cache entries always used the global `resolver-cache-config` TTL, even when a resolver ConfigMap set a positive TTL. Entries now use the resolver-specific TTL when it is greater than 0 and fall back to the global TTL otherwise.

> * :bug: fix: respect per-resolver TTL override in cache (#9625)
> Resolver cache now respects per-resolver TTL values set in individual
> resolver ConfigMaps (e.g., bundleresolver-config, git-resolver-config),
> instead of always using the global resolver-cache-config TTL.

### Premature PipelineRun failures after pod eviction fixed

- Affects: PipelineRuns with pod-backed TaskRuns

During the race after pod eviction and before the TaskRun status updates to Running, a stale cached failure could permanently fail the PipelineRun even if the TaskRun recovered. For cached `PodEvicted` failures, the controller now makes one API-server GET per failed TaskRun per reconcile and uses the fresh status; if verification fails for a reason other than NotFound, reconciliation returns an error.

> * :bug: Fix PipelineRun premature failure when TaskRun recovers after pod eviction (#9640)

### Affinity assistant volume names now fit Kubernetes’ limit

- Affects: affinity assistants in `AffinityAssistantPerPipelineRun` or `AffinityAssistantPerPipelineRunWithIsolation` mode with workspace `volumeClaimTemplate` names that generate volume names longer than 63 characters (Inferred)

A long `volumeClaimTemplate.metadata.name` could make the generated volume name exceed Kubernetes’ 63-character limit, causing the affinity assistant pod to fail creation and block the pipeline. Overlong names now receive a hash suffix when truncated; shorter names remain unchanged.

> * :bug: fix: truncate affinity assistant volume names to 63 characters (#9752)
> Affinity assistant StatefulSet no longer fails when workspace volumeClaimTemplate names exceed 63 characters. Long volume names are now automatically truncated with a hash suffix to stay within the Kubernetes limit.

### PipelineRun timeouts now allow finally tasks to run

- Affects: PipelineRuns with finally tasks

When either an explicit `timeouts.tasks` timeout or the calculated `timeouts.pipeline - timeouts.finally` timeout was exceeded while finally tasks remained, the PipelineRun was marked Failed immediately, so those tasks did not run. It now remains running until pending finally tasks complete, then fails with the timeout reason.

> * :bug: fix: allow finally tasks to run when tasks timeout is exceeded (#9709)
> Fix a bug where finally tasks were not executed when the tasks timeout
> (either explicit via `timeouts.tasks` or calculated as `timeouts.pipeline -
> timeouts.finally`) was exceeded. The PipelineRun was immediately marked as
> Failed without giving finally tasks a chance to run. Now the pipeline
> continues running with reason `PipelineRunTimeoutRunningFinally` until
> finally tasks complete.

### Resolver framework timeout goroutine leaks fixed

- Affects: setups using the local or remote resolver framework reconcilers

When `Resolve()` outlasted its deadline, the caller returned while the goroutine blocked sending on `errChan` or `resourceChan`, leaking a goroutine and held resources; sustained load could degrade the controller until restart. The buffered channels now let the goroutine send its result and exit.

> * :bug: fix: resolve goroutine leak from unbuffered channels in resolver reconciler (#10098)

## Enhancements (1)

- Optional termination message compression (To use: The alpha feature flag `enable-termination-message-compression` is required to enable termination message compression.)

## Dependency updates (32)

32 more, none mentioning security

- `* :hammer: chore: bump OpenTelemetry semconv to match SDK version (#9697)`
- `* :hammer: build(deps): bump the all group in /tekton with 4 updates (#10126)`
- `* :hammer: build(deps): bump golangci/golangci-lint-action from 9.2.0 to 9.2.1 (#10125)`
- `* :hammer: build(deps): bump go.opentelemetry.io/otel/trace from 1.43.0 to 1.44.0 (#10118)`
- `* :hammer: build(deps): bump golang.org/x/crypto from 0.51.0 to 0.52.0 (#10095)`
- `* :hammer: build(deps): bump the all group in /tekton with 4 updates (#10085)`
- `* :hammer: build(deps): bump github/codeql-action from 4.35.4 to 4.35.5 (#10084)`
- `* :hammer: build(deps): bump step-security/harden-runner from 2.19.2 to 2.19.4 (#10083)`
- `* :hammer: build(deps): bump zizmorcore/zizmor-action from 0.5.3 to 0.5.6 (#10082)`
- `* :hammer: build(deps): bump github.com/sigstore/sigstore/pkg/signature/kms/gcp from 1.10.5 to 1.10.6 (#10081)`
- `* :hammer: build(deps): bump github.com/spiffe/spire-api-sdk from 1.14.6 to 1.15.0 (#10080)`
- `* :hammer: build(deps): bump github.com/sigstore/sigstore/pkg/signature/kms/aws from 1.10.5 to 1.10.6 (#10060)`
- `* :hammer: build(deps): bump github.com/google/go-containerregistry from 0.21.5 to 0.21.6 (#10058)`
- `* :hammer: build(deps): bump github.com/sigstore/sigstore from 1.10.5 to 1.10.6 (#10057)`
- `* :hammer: build(deps): bump github.com/sigstore/sigstore/pkg/signature/kms/hashivault from 1.10.5 to 1.10.6 (#10047)`
- `* :hammer: build(deps): bump the all group in /tekton with 4 updates (#10039)`
- `* :hammer: build(deps): bump step-security/harden-runner from 2.19.1 to 2.19.2 (#10038)`
- `* :hammer: build(deps): bump actions/dependency-review-action from 4.9.0 to 5.0.0 (#10037)`
- `* :hammer: build(deps): bump google.golang.org/grpc from 1.81.0 to 1.81.1 (#10036)`
- `* :hammer: build(deps): bump k8s.io/apiextensions-apiserver from 0.35.4 to 0.35.5 (#10033)`
- `* :hammer: build(deps): bump github.com/jenkins-x/go-scm from 1.15.21 to 1.15.22 (#10032)`
- `* :hammer: build(deps): bump github.com/google/cel-go from 0.28.0 to 0.28.1 (#10012)`
- `* :hammer: build(deps): bump k8s.io/client-go from 0.35.4 to 0.35.5 (#10011)`
- `* :hammer: build(deps): bump k8s.io/code-generator from 0.35.4 to 0.35.5 (#10010)`
- `* :hammer: build(deps): bump k8s.io/client-go from 0.35.4 to 0.35.5 in /test/custom-task-ctrls/wait-task-beta (#10009)`
- `* :hammer: build(deps): bump golang.org/x/crypto from 0.50.0 to 0.51.0 (#10005)`
- `* :hammer: build(deps): bump github/codeql-action from 4.35.2 to 4.35.4 (#9992)`
- `` * :hammer: build(deps): bump chainguard/go from `7ec9277` to `a4477c3` in /tekton in the all group (#9991) ``
- `* :hammer: build(deps): bump step-security/harden-runner from 2.19.0 to 2.19.1 (#9978)`
- `* :hammer: build(deps): bump chainguard-dev/actions from 1.6.17 to 1.6.19 (#9977)`
- `* :hammer: build(deps): bump the all group in /tekton with 4 updates (#9975)`
- `* :hammer: build(deps): bump google.golang.org/grpc from 1.80.0 to 1.81.0 (#9954)`

## Other changes (17)

- Tracing guidance on sensitive data in exported traces
- Resolver framework skips re-resolution for `ResolutionRequest`s with data
- Fetchable module requirement for `gen-crd-api-reference-docs`
- Dependabot regeneration commits use a CLA-authorized identity and DCO sign-off
- `dependabot-regen` workflow uses `secrets.CHATOPS_TOKEN` for checkout and pushes
- Go version directive and test submodules updated to 1.26
- TaskRun native-sidecar detection cached per reconciler
- Reduced CI e2e matrix for draft PRs
- `.github/dependabot.yml` regenerated from its source configuration
- Pipeline dependency updated to 1.12.0 in `wait-task-beta` tests
- `actions/setup-go` updated from 5.2.0 to 6.4.0
- Build-image and tool references in `/tekton` updated
- `setup-go` version comment fixed in `codeql-analysis.yml`
- Release cheat sheet update for website sync configuration
- `releases.md` updated with v1.12.0 LTS and v1.11 and v1.9 entries
- Retry-count substitution clarified for inline task specs
- Testing best practices added

---

Summaries are produced by AI, so they can be wrong, and they are provided without warranty.
