Change runner to connect to otel collector on pre-job

charmcraft.yaml

@@ -240,6 +240,12 @@ config:
       description: >-
         The log level for the runner manager application. The value can be CRITICAL, FATAL, ERROR,
         WARNING, INFO, or DEBUG.
+    otel-collector-endpoint:
+      type: string
+      default: ""
+      description: >-
+        The endpoint to send OpenTelemetry metrics to in the format "host:port". If not set, OpenTelemetry
+        will be disabled.
 
 actions:
   check-runners:

docs/changelog.md

@@ -2,6 +2,10 @@
 
 This changelog documents user-relevant changes to the GitHub runner charm.
 
+## 2026-04-27
+
+- Added configuration option `otel-collector-endpoint` to enable the otel-collector to export metric. Setting this configuration option will add the environment variable ACTION_OTEL_EXPORTER_OTLP_ENDPOINT to the runner, which allow users to configure their own metrics to be exported.
+
 ## 2026-04-22
 
 - Removed `KillMode=process` from the runner manager systemd service, restoring the default `control-group` kill mode. This ensures all child processes in the service's cgroup are properly terminated when the service stops, preventing orphaned runner processes.

github-runner-manager/src/github_runner_manager/configuration/base.py

@@ -91,6 +91,7 @@ class SupportServiceConfig(BaseModel):
         aproxy_redirect_ports: A list of ports to redirect to the aproxy proxy.
         dockerhub_mirror: The dockerhub mirror to use for runners.
         ssh_debug_connections: The information on the ssh debug services.
+        otel_collector_config: The configuration for the OpenTelemetry collector.
         custom_pre_job_script: The custom pre-job script to run before the job.
     """
 
@@ -103,6 +104,7 @@ class SupportServiceConfig(BaseModel):
     dockerhub_mirror: str | None
     ssh_debug_connections: "list[SSHDebugConnection]"
     custom_pre_job_script: str | None
+    otel_collector_config: Optional["OtelCollectorConfig"] = None
 
     @root_validator(pre=False, skip_on_failure=True)
     @classmethod
@@ -127,6 +129,18 @@ def check_use_aproxy(cls, values: dict) -> dict:
         return values
 
 
+class OtelCollectorConfig(BaseModel):
+    """Configuration for OpenTelemetry collector.
+
+    Attributes:
+        host: The OpenTelemetry collector hostname.
+        port: The OpenTelemetry collector port.
+    """
+
+    host: str
+    port: int = Field(gt=0, le=65535)
+
+
 class ProxyConfig(BaseModel):
     """Proxy configuration.
 

github-runner-manager/src/github_runner_manager/openstack_cloud/openstack_runner_manager.py

@@ -170,18 +170,26 @@ def _generate_cloud_init(self, runner_context: RunnerContext) -> str:
             if service_config.ssh_debug_connections
             else None
         )
+        otel_collector_config = service_config.otel_collector_config
+        otel_collector_endpoint = (
+            f"{otel_collector_config.host}:{otel_collector_config.port}"
+            if otel_collector_config
+            else ""
+        )
         env_contents = jinja.get_template("env.j2").render(
             pre_job_script=str(PRE_JOB_SCRIPT),
             dockerhub_mirror=service_config.dockerhub_mirror or "",
             ssh_debug_info=ssh_debug_info,
             tmate_server_proxy=runner_http_proxy,
+            otel_collector_endpoint=otel_collector_endpoint,
         )
         pre_job_contents_dict = {
             "issue_metrics": True,
             "metrics_exchange_path": str(METRICS_EXCHANGE_PATH),
             "do_repo_policy_check": False,
             "custom_pre_job_script": service_config.custom_pre_job_script,
             "allow_external_contributor": self._config.allow_external_contributor,
+            "otel_collector_endpoint": otel_collector_endpoint,
         }
 
         pre_job_contents = jinja.get_template("pre-job.j2").render(pre_job_contents_dict)

github-runner-manager/src/github_runner_manager/templates/env.j2

@@ -15,3 +15,6 @@ TMATE_SERVER_HOST={{ssh_debug_info.local_proxy_host}}
 TMATE_SERVER_PORT={{ssh_debug_info.local_proxy_port}}
 {% endif %}
 {% endif %}
+{% if otel_collector_endpoint %}
+ACTION_OTEL_EXPORTER_OTLP_ENDPOINT={{otel_collector_endpoint}}
+{% endif %}

github-runner-manager/src/github_runner_manager/templates/pre-job.j2

@@ -131,6 +131,57 @@ jq -n \
   logger -s "Contributor check passed - proceeding to execute jobs"
 {% endif %}
 
+# Setup the OpenTelemetry collector configurations.
+{% if otel_collector_endpoint %}
+/usr/bin/logger -s "OpenTelemetry collector is enabled."
+/usr/bin/logger -s "Additional OpenTelemetery collector configuration can be added."
+/usr/bin/logger -s "The exporter endpoint is at the environment variable ACTION_OTEL_EXPORTER_OTLP_ENDPOINT."
+/usr/bin/sudo /usr/bin/mkdir -p /etc/otelcol/config.d
+/usr/bin/sudo /usr/bin/touch /etc/otelcol/config.d/github.yaml
+/usr/bin/sudo /usr/bin/tee /etc/otelcol/config.d/github.yaml <<EOF
+receivers:
+  hostmetrics:
+    collection_interval: 10s
+    scrapers:
+      cpu:
+      memory:
+      disk:
+      filesystem:
+      network:
+      load:
+processors:
+  attributes/github_labels:
+    actions:
+      - key: github_runner
+        action: upsert
+        value:  "$RUNNER_NAME"
+      - key: github_workflow
+        action: upsert
+        value:  "$GITHUB_WORKFLOW"
+      - key: github_job
+        action: upsert
+        value:  "$GITHUB_JOB"
+      - key: github_repository
+        action: upsert
+        value:  "$GITHUB_REPOSITORY"
+  batch:
+exporters:
+  otlp/mimir:
+    endpoint: {{ otel_collector_endpoint }}
+    tls:
+      insecure: true
+service:
+  pipelines:
+    metrics:
+      receivers: [hostmetrics]
+      processors: [attributes/github_labels, batch]
+      exporters: [otlp/mimir]
+EOF
+
+/usr/bin/sudo /usr/bin/snap enable opentelemetry-collector
+/usr/bin/sudo /usr/bin/snap start opentelemetry-collector
+{% endif %}
+
 if [[ -n "$DOCKERHUB_MIRROR" ]]; then
   logger -s "A private docker registry is setup as a dockerhub mirror for this self-hosted runner."
   logger -s "The docker daemon on this self-hosted runner is configured to use the dockerhub mirror."

src/charm_state.py

@@ -17,6 +17,7 @@
 
 import yaml
 from github_runner_manager.configuration import ProxyConfig, SSHDebugConnection
+from github_runner_manager.configuration.base import OtelCollectorConfig
 from github_runner_manager.configuration.github import (
     GitHubAppAuth,
     GitHubAuth,
@@ -69,6 +70,7 @@
 VIRTUAL_MACHINES_CONFIG_NAME = "virtual-machines"
 CUSTOM_PRE_JOB_SCRIPT_CONFIG_NAME = "pre-job-script"
 RUNNER_MANAGER_LOG_LEVEL_CONFIG_NAME = "runner-manager-log-level"
+OTEL_COLLECTOR_ENDPOINT_CONFIG_NAME = "otel-collector-endpoint"
 
 # Integration names
 COS_AGENT_INTEGRATION_NAME = "cos-agent"
@@ -840,6 +842,39 @@ def _build_ssh_debug_connection_from_charm(charm: CharmBase) -> list[SSHDebugCon
     return ssh_debug_connections
 
 
+def _build_otel_collector_config_from_charm(charm: CharmBase) -> OtelCollectorConfig | None:
+    """Initialize the OtelCollectorConfig from charm configuration.
+
+    Args:
+        charm: The charm instance.
+
+    Returns:
+        OtelCollectorConfig if endpoint config is set; otherwise None.
+    """
+    endpoint = cast(str, charm.config.get(OTEL_COLLECTOR_ENDPOINT_CONFIG_NAME, ""))
+    if not endpoint:
+        return None
+
+    parsed_endpoint = urlsplit(f"//{endpoint}")
+    if not parsed_endpoint.hostname or parsed_endpoint.port is None:
+        raise CharmConfigInvalidError(
+            f"Invalid {OTEL_COLLECTOR_ENDPOINT_CONFIG_NAME} config, expected host:port"
+        )
+
+    if (
+        parsed_endpoint.username
+        or parsed_endpoint.password
+        or parsed_endpoint.path
+        or parsed_endpoint.query
+        or parsed_endpoint.fragment
+    ):
+        raise CharmConfigInvalidError(
+            f"Invalid {OTEL_COLLECTOR_ENDPOINT_CONFIG_NAME} config, expected host:port"
+        )
+
+    return OtelCollectorConfig(host=parsed_endpoint.hostname, port=parsed_endpoint.port)
+
+
 def _build_planner_config_from_charm(charm: CharmBase) -> PlannerConfig | None:
     """Initialize planner endpoint and token from relation data.
 
@@ -896,6 +931,7 @@ class CharmState:  # pylint: disable=too-many-instance-attributes
         runner_proxy_config: Proxy-related configuration for the runner.
         runner_config: The charm configuration related to runner VM configuration.
         ssh_debug_connections: SSH debug connections configuration information.
+        otel_collector_config: OpenTelemetry collector configuration information.
         planner_config: Planner endpoint and token from relation data.
     """
 
@@ -905,6 +941,7 @@ class CharmState:  # pylint: disable=too-many-instance-attributes
     charm_config: CharmConfig
     runner_config: OpenstackRunnerConfig
     ssh_debug_connections: list[SSHDebugConnection]
+    otel_collector_config: OtelCollectorConfig | None
     planner_config: PlannerConfig | None
 
     @classmethod
@@ -923,6 +960,11 @@ def _store_state(cls, state: "CharmState") -> None:
         state_dict["ssh_debug_connections"] = [
             debug_info.json() for debug_info in state_dict["ssh_debug_connections"]
         ]
+        state_dict["otel_collector_config"] = (
+            json.loads(state_dict["otel_collector_config"].json())
+            if state_dict["otel_collector_config"]
+            else None
+        )
         json_data = json.dumps(state_dict, ensure_ascii=False)
         CHARM_STATE_PATH.write_text(json_data, encoding="utf-8")
 
@@ -975,6 +1017,12 @@ def from_charm(cls, charm: CharmBase) -> "CharmState":  # noqa: C901
             logger.error("Invalid SSH debug info: %s.", exc)
             raise CharmConfigInvalidError("Invalid SSH Debug info") from exc
 
+        try:
+            otel_collector_config = _build_otel_collector_config_from_charm(charm)
+        except (ValidationError, ValueError) as exc:
+            logger.error("Invalid OpenTelemetry collector config: %s.", exc)
+            raise CharmConfigInvalidError("Invalid OpenTelemetry collector config") from exc
+
         planner_config = _build_planner_config_from_charm(charm)
 
         state = cls(
@@ -984,6 +1032,7 @@ def from_charm(cls, charm: CharmBase) -> "CharmState":  # noqa: C901
             charm_config=charm_config,
             runner_config=runner_config,
             ssh_debug_connections=ssh_debug_connections,
+            otel_collector_config=otel_collector_config,
             planner_config=planner_config,
         )
 

src/factories.py

@@ -53,6 +53,7 @@ def create_application_configuration(
         runner_proxy_config=state.runner_proxy_config,
         dockerhub_mirror=state.charm_config.dockerhub_mirror,
         ssh_debug_connections=state.ssh_debug_connections,
+        otel_collector_config=state.otel_collector_config,
         use_aproxy=state.charm_config.use_aproxy,
         aproxy_exclude_addresses=state.charm_config.aproxy_exclude_addresses,
         aproxy_redirect_ports=state.charm_config.aproxy_redirect_ports,

tests/integration/helpers/openstack.py

@@ -125,10 +125,11 @@ def run_in_instance(
         exit_code, _, _ = run_in_unit(self.juju, unit_name, f"ls {key_path}")
         assert exit_code == 0, f"Unable to find key file {key_path}"
         ssh_cmd = f'ssh -i {key_path} -o "StrictHostKeyChecking no" ubuntu@{ip} {command}'
-        ssh_cmd_as_ubuntu_user = f"su - ubuntu -c '{ssh_cmd}'"
-        logging.warning("ssh_cmd: %s", ssh_cmd_as_ubuntu_user)
+        # The SSH command needs to be run as the manager user to have access to the SSH keys.
+        ssh_cmd_as_manager_user = f"su - {constants.RUNNER_MANAGER_USER} -c '{ssh_cmd}'"
+        logging.warning("ssh_cmd: %s", ssh_cmd_as_manager_user)
         exit_code, stdout, stderr = run_in_unit(
-            self.juju, unit_name, ssh_cmd_as_ubuntu_user, timeout
+            self.juju, unit_name, ssh_cmd_as_manager_user, timeout
         )
         logger.info(
             "Run command '%s' in runner with result %s: '%s' '%s'",

tests/integration/test_charm_runner.py

@@ -10,7 +10,11 @@
 from github.Branch import Branch
 from github.Repository import Repository
 
-from charm_state import BASE_VIRTUAL_MACHINES_CONFIG_NAME, CUSTOM_PRE_JOB_SCRIPT_CONFIG_NAME
+from charm_state import (
+    BASE_VIRTUAL_MACHINES_CONFIG_NAME,
+    CUSTOM_PRE_JOB_SCRIPT_CONFIG_NAME,
+    OTEL_COLLECTOR_ENDPOINT_CONFIG_NAME,
+)
 from tests.integration.helpers.common import (
     DISPATCH_TEST_WORKFLOW_FILENAME,
     DISPATCH_WAIT_TEST_WORKFLOW_FILENAME,
@@ -186,3 +190,47 @@ def test_custom_pre_job_script(
     logs = get_job_logs(workflow_run.jobs("latest")[0])
     assert "SSH config" in logs
     assert "proxycommand socat - PROXY:squid.internal:%h:%p,proxyport=3128" in logs
+
+
+@pytest.mark.openstack
+@pytest.mark.abort_on_fail
+def test_otel_collector_endpoint_pre_job_installs_config(
+    juju: jubilant.Juju,
+    app: str,
+    github_repository: Repository,
+    test_github_branch: Branch,
+    instance_helper: OpenStackInstanceHelper,
+) -> None:
+    """
+    arrange: A working application with one runner and otel collector endpoint configured.
+    act: Dispatch a workflow to run pre-job script.
+    assert: The workflow writes otel collector config to /etc/otelcol/config.d/github.yaml.
+    """
+    endpoint = "10.10.0.12:4317"
+    juju.config(
+        app,
+        values={
+            BASE_VIRTUAL_MACHINES_CONFIG_NAME: "1",
+            OTEL_COLLECTOR_ENDPOINT_CONFIG_NAME: endpoint,
+        },
+    )
+    wait_for_runner_ready(juju, app)
+
+    dispatch_workflow(
+        app_name=app,
+        branch=test_github_branch,
+        github_repository=github_repository,
+        conclusion="success",
+        workflow_id_or_name=DISPATCH_TEST_WORKFLOW_FILENAME,
+        dispatch_input={"runner": app},
+    )
+
+    exit_code, stdout, stderr = instance_helper.run_in_instance(
+        unit_name=f"{app}/0",
+        command="sudo cat /etc/otelcol/config.d/github.yaml",
+    )
+
+    assert exit_code == 0, stderr
+    assert stdout is not None
+    assert "exporters:" in stdout
+    assert f"endpoint: {endpoint}" in stdout

tests/unit/conftest.py

@@ -188,5 +188,6 @@ def complete_charm_state_fixture():
                 ed25519_fingerprint="SHA256:ed25519",
             ),
         ],
+        otel_collector_config=None,
         planner_config=None,
     )