Odigos Files

Changes by Kind

Uncategorized

• .NET instrumentation artifacts downloaded during the odiglet build are now verified against pinned SHA256 checksums. (#5515, @aviadhahami)
• AWS connector docs (#5661, @TheErikNaz)
• Add a URL templatization configuration page to the UI for managing default rules and custom URL templates. (#5600, @alonkeyval)
• Added public documentation for verifying Odigos image signatures and SBOM attestations. (#5585, @aviadhahami)
• All published odigos container images are now signed with cosign (keyless) and carry CycloneDX SBOM attestations, and signatures follow images across registry mirrors and RC promotion. (#5516, @aviadhahami)
• Bumped golang.org/x/mod to v0.40.0 in the odiglet module. (#5574, @aviadhahami)
• Bumped golang.org/x/net and golang.org/x/text in the device plugin. (#5548, @aviadhahami)
• Bumped grpc_health_probe to v0.4.55. (#5575, @aviadhahami)
• Bumped the Go toolchain to 1.26.6. (#5556, @aviadhahami)
• Bumped the OpenTelemetry Java agent to v2.20.0, resolving jackson-databind and jackson-core vulnerabilities. (#5510, @aviadhahami)
• Chore(deps): update @odigos/ui-kit to version 0.0.295 (#5733, @octo-sts[bot])
• Collector version v0.159.0 (#5714, @TheErikNaz)
• Documented Cloud Connectors for Odigos Central (enablement, AWS onboarding, destinations, instrumentation, and connector lifecycle) (#5431, @rayray221)
• Documented VM Agent application log collection (LOGS), including per-source collection methods. (#5544, @rayray221)
• EBPF instrumentations now report telemetry.distro.name with the Odigos distribution name (for example java-ebpf-instrumentations, golang-enterprise). (#5731, @octo-sts[bot])
• Feat(destinations): support mTLS in OTLP destinations (#5639, @damemi)
• Feat: Odigos Insights (#5722, @BenElferink)
• Fix OpenShift DeploymentConfig workload identity in the collector by preferring Odigos workload attributes over shared Kubernetes resource attributes. (#5601, @blumamir)
• Fix UI RBAC so recommendation dismiss/update works. (#5687, @blumamir)
• Fix overall workload Odigos health status double-counting auto-rollback when agent injection enabled already reflects it. (#5688, @blumamir)
• Fix pods manifest injection status reporting for CronJob workloads. (#5642, @blumamir)
• Fix the Kafka destination never delivering telemetry: the broker list was rendered as a string instead of a list, and all signals shared a single exporter so they all exported to the logs topic when no topic was configured. (#5693, @cursor[bot])
• Fix(helm): stop Insights install from hanging when ClickHouse storage cannot bind (#5730, @octo-sts[bot])
• Fix: bump UI Docker/CI Node to Active LTS 24 (#5699, @BenElferink)
• Fix: copy profiles dictionary on datastream fan-out so destination export works (#5724, @tamirdavid1)
• Fix: improve concurrency for diagnose workloads on large clusters (#5598, @damemi)
• Fix: node collectors now route traces to the gateway by trace id when tail sampling or service I/O trace correlations are active, not only when the service graph is enabled. Previously a cluster with the service graph disabled and a multi-replica gateway made tail sampling decisions on partial traces. (#5662, @cursor[bot])
• Fixed PII masking customFormatMaskings masking for values with spaces (#5552, @cursor[bot])
• Fixed the Java agent jar being rewritten in place during an Odigos upgrade, which broke instrumentation in already-running JVMs whenever the agent version changed. Running pods now keep their existing agent until they restart; newly started pods pick up the new one. (#5549, @aviadhahami)
• Fixed the Network Metrics instrumentation rule being created as an unknown, unopenable rule when its "Enable network metrics" toggle was turned off. The rule is now enabled by its existence and disabled with the standard rule disable control. (#5606, @alonkeyval)
• Fixed the cluster collector crash-looping when Odigos own metrics were routed only to metrics destinations and no metrics destination could be configured (for example a destination missing a required field), which stopped all telemetry instead of only that destination. (#5645, @cursor[bot])
• Fixed the vulnerability scan workflow silently discarding results for large images, and made it fail the run when CRITICAL vulnerabilities are found. (#5508, @aviadhahami)
• IgnoredContainers and ignoredNamespaces set via Helm or config now replace their default lists instead of merging with them, so users can opt in to instrumenting previously always-ignored sidecars and namespaces. OpenShift system namespaces and the Odigos install namespace are still handled separately. (#5716, @blumamir)
• OSS: Switch to docker.io/keyval as default container image registry (#5532, @damemi)
• Pass additional metrics to VictoriaMetrics (odigos own metrics) including ebpf ringbuffer state and cpu/memory (#5554, @natyharary)
• Sampling HTTP route and routePrefix rules now match templated path segments (* and {name}), not only literal strings. (#5535, @blumamir)
• Span metrics are now enabled by default for all metrics destinations. Use collectSpanMetrics: false to opt out. (#5607, @blumamir)