From d326f9cf028e7fc43350972cea99ffa28cf81cec Mon Sep 17 00:00:00 2001 From: Benedikt Bongartz Date: Tue, 15 Apr 2025 23:16:30 +0200 Subject: [PATCH] k8s: configure otel sdk via instrumentation cr Signed-off-by: Benedikt Bongartz --- kubernetes/llama-serve/granite-8b/vllm.yaml | 2 ++ kubernetes/llama-serve/llama3.1-70b/vllm.yaml | 2 ++ kubernetes/llama-serve/llama3.2-3b/vllm.yaml | 2 ++ kubernetes/llama-serve/watt-8b/vllm.yaml | 2 ++ kubernetes/llama-stack/deployment.yaml | 6 +----- .../observability/otel-sdk-hub-config.yaml | 19 +++++++++++++++++++ 6 files changed, 28 insertions(+), 5 deletions(-) create mode 100644 kubernetes/observability/otel-sdk-hub-config.yaml diff --git a/kubernetes/llama-serve/granite-8b/vllm.yaml b/kubernetes/llama-serve/granite-8b/vllm.yaml index b607cd48d..284bcda15 100644 --- a/kubernetes/llama-serve/granite-8b/vllm.yaml +++ b/kubernetes/llama-serve/granite-8b/vllm.yaml @@ -20,6 +20,8 @@ spec: creationTimestamp: null labels: app: granite-8b + annotations: + instrumentation.opentelemetry.io/inject-sdk: "hub" # NOTE: May extend with "." spec: containers: - args: diff --git a/kubernetes/llama-serve/llama3.1-70b/vllm.yaml b/kubernetes/llama-serve/llama3.1-70b/vllm.yaml index 712f78a78..9a3b3672f 100644 --- a/kubernetes/llama-serve/llama3.1-70b/vllm.yaml +++ b/kubernetes/llama-serve/llama3.1-70b/vllm.yaml @@ -14,6 +14,8 @@ spec: creationTimestamp: null labels: app: llama31-70b + annotations: + instrumentation.opentelemetry.io/inject-sdk: "hub" # NOTE: May extend with "." spec: volumes: - name: hf-cache diff --git a/kubernetes/llama-serve/llama3.2-3b/vllm.yaml b/kubernetes/llama-serve/llama3.2-3b/vllm.yaml index 7d8c036d0..467200936 100644 --- a/kubernetes/llama-serve/llama3.2-3b/vllm.yaml +++ b/kubernetes/llama-serve/llama3.2-3b/vllm.yaml @@ -20,6 +20,8 @@ spec: creationTimestamp: null labels: app: llama32-3b + annotations: + instrumentation.opentelemetry.io/inject-sdk: "hub" # NOTE: May extend with "." spec: containers: - args: diff --git a/kubernetes/llama-serve/watt-8b/vllm.yaml b/kubernetes/llama-serve/watt-8b/vllm.yaml index f65705d2b..19b5a0ead 100644 --- a/kubernetes/llama-serve/watt-8b/vllm.yaml +++ b/kubernetes/llama-serve/watt-8b/vllm.yaml @@ -14,6 +14,8 @@ spec: creationTimestamp: null labels: app: watt-8b + annotations: + instrumentation.opentelemetry.io/inject-sdk: "hub" # NOTE: May extend with "." spec: volumes: - name: hf-cache diff --git a/kubernetes/llama-stack/deployment.yaml b/kubernetes/llama-stack/deployment.yaml index eda938446..f7e9fbee0 100644 --- a/kubernetes/llama-stack/deployment.yaml +++ b/kubernetes/llama-stack/deployment.yaml @@ -9,7 +9,7 @@ spec: template: metadata: annotations: - sidecar.opentelemetry.io/inject: llamastack-otelsidecar + instrumentation.opentelemetry.io/inject-sdk: "hub" # NOTE: May extend with "." labels: app: llamastack spec: @@ -44,10 +44,6 @@ spec: value: 'https://llama32-3b-llama-serve.apps.ocp-beta-test.nerc.mghpcc.org/v1' - name: VLLM_API_TOKEN value: fake - - name: OTEL_SERVICE_NAME - value: om-llamastack - - name: OTEL_TRACE_ENDPOINT - value: 'http://localhost:4318/v1/traces' - name: SAFETY_MODEL value: meta-llama/Llama-Guard-3-8B - name: SAFETY_VLLM_URL diff --git a/kubernetes/observability/otel-sdk-hub-config.yaml b/kubernetes/observability/otel-sdk-hub-config.yaml new file mode 100644 index 000000000..f8590abdd --- /dev/null +++ b/kubernetes/observability/otel-sdk-hub-config.yaml @@ -0,0 +1,19 @@ +apiVersion: opentelemetry.io/v1alpha1 +kind: Instrumentation +metadata: + name: hub +spec: + exporter: + endpoint: http://otel-collector-collector.observability-hub.svc.cluster.local:4317 + env: + # NOTE: can this be provided by the service itself? + # - name: "OTEL_SERVICE_NAME" + # value: "om-llamastack" + - name: "OTEL_EXPORTER_OTLP_TRACES_INSECURE" + value: "true" + # NOTE: Is http requiered? + - name: "OTEL_TRACE_ENDPOINT" + value: "http://otel-collector-collector.observability-hub.svc.cluster.local:4318/v1/traces" + propagators: + - tracecontext + - baggage