scylla: add container-based config discovery support (DataDog#24498)

* Add container-based config discovery support to scylla

Adds a discovery strategy targeting the openmetrics_endpoint on port 9180
(scylla's Prometheus metrics port), generates auto_conf.yaml with
ad_identifiers=[scylla], wires get_e2e_discovery_metadata() into the E2E
fixture, and adds test_e2e_discovery/test_e2e_discovery_all_candidates
tests to validate discovered configs against the real scylla-db container.

Environment: Datadog workspace

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

* Add changelog entry for DataDog#24498

Environment: Datadog workspace

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

* Tighten discovery stability log-pattern regex

Anchor the benign-error exclusion to the specific "rpc - client <addr>: server
connection dropped: connection is closed" log shape instead of a bare trailing
substring match, per Codex review feedback, so an unrelated error line that
happens to contain the same trailing text isn't silently excluded too.

Environment: Datadog workspace

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

* cleanup

---------

Co-authored-by: Claude Sonnet 5 <noreply@anthropic.com>

Author: vitkyrka

scylla/assets/configuration/spec.yaml

@@ -2,6 +2,12 @@ name: Scylla
 fleet_configurable: true
 files:
 - name: scylla.yaml
+  discovery:
+    strategies:
+    - template: discovery/openmetrics_from_ports
+      overrides:
+        port_hints:
+        - 9180
   options:
   - template: init_config
     options:
@@ -64,3 +70,10 @@ files:
       - type: multi_line
         pattern: \d{4}\-(0?[1-9]|1[012])\-(0?[1-9]|[12][0-9]|3[01])
         name: new_log_start_with_date
+- name: auto_conf.yaml
+  options:
+  - template: ad_identifiers
+    overrides:
+      value.example:
+      - scylla
+  - template: auto_conf/discovery

scylla/changelog.d/24498.added

@@ -0,0 +1 @@
+Add container-based config discovery support.

scylla/datadog_checks/scylla/config_models/discovery.py

@@ -0,0 +1,42 @@
+# (C) Datadog, Inc. 2026-present
+# All rights reserved
+# Licensed under a 3-clause BSD style license (see LICENSE)
+
+# This file is autogenerated.
+# To change this file you should edit assets/configuration/spec.yaml and then run the following commands:
+#     ddev -x validate config -s <INTEGRATION_NAME>
+#     ddev -x validate models -s <INTEGRATION_NAME>
+
+from __future__ import annotations
+
+from collections.abc import Iterator
+from typing import Any
+
+from datadog_checks.base.utils.discovery import Service, candidate_ports
+from datadog_checks.scylla.config_models import discovery_overrides
+from datadog_checks.scylla.config_models.instance import InstanceConfig
+from datadog_checks.scylla.config_models.shared import SharedConfig
+
+
+def _generated_candidates(service: Service) -> Iterator[dict[str, Any]]:
+    shared = SharedConfig.model_validate({}, context={'configured_fields': frozenset()}).model_dump(
+        by_alias=True, mode='json', exclude_none=True
+    )
+    # discovery[0]: from_ports
+    for port in candidate_ports(service, [9180]):
+        ctx = {'port': port}
+        instance_data = {
+            'openmetrics_endpoint': 'http://{service.host}:{port.number}/metrics'.format(service=service, **ctx),
+        }
+        instance = InstanceConfig.model_validate(
+            instance_data, context={'configured_fields': frozenset(instance_data)}
+        ).model_dump(by_alias=True, mode='json', exclude_none=True)
+        yield {'init_config': shared, 'instances': [instance]}
+
+
+def candidates(service: Service) -> Iterator[dict[str, Any]]:
+    override = getattr(discovery_overrides, 'candidates', None)
+    if override is None:
+        yield from _generated_candidates(service)
+    else:
+        yield from override(service, default=_generated_candidates)

scylla/datadog_checks/scylla/config_models/discovery_overrides.py

@@ -0,0 +1,12 @@
+# (C) Datadog, Inc. 2026-present
+# All rights reserved
+# Licensed under a 3-clause BSD style license (see LICENSE)
+
+# Override the generated discovery candidates() for this integration.
+#
+# Define a candidates(service, default) function to wrap or replace the generated
+# candidate generation. `default` is the generated generator; call it to reuse
+# the spec-driven candidates, or ignore it to replace them entirely.
+#
+# def candidates(service, default):
+#     yield from default(service)

scylla/datadog_checks/scylla/config_models/discovery_strategies.py

@@ -0,0 +1,18 @@
+# (C) Datadog, Inc. 2026-present
+# All rights reserved
+# Licensed under a 3-clause BSD style license (see LICENSE)
+
+# Here you can define custom (local:) discovery strategies for this integration.
+#
+# Decorate a generator with @discovery_strategy (imported from
+# datadog_checks.base.utils.discovery) and reference it from the spec discovery
+# stanza as `strategy: local:<function_name>`. The function receives the
+# discovered Service plus the inputs declared in the spec and yields one context
+# (ctx) mapping per candidate, exposing the keys listed in `provides`.
+#
+# from datadog_checks.base.utils.discovery import discovery_strategy
+#
+# @discovery_strategy(provides=('svc',))
+# def from_some_config(service, config_path):
+#     ...
+#     yield {'svc': ...}

scylla/datadog_checks/scylla/data/auto_conf.yaml

@@ -0,0 +1,19 @@
+## @param ad_identifiers - list of strings - required
+## A list of container identifiers that are used by Autodiscovery to identify
+## which container the check should be run against. For more information, see:
+## https://docs.datadoghq.com/agent/guide/ad_identifiers/
+#
+ad_identifiers:
+  - scylla
+
+## Enables configuration discovery
+#
+discovery: {}
+
+## Unused init configuration
+#
+init_config:
+
+## Unused instance configuration
+#
+instances: []

scylla/tests/conftest.py

@@ -6,7 +6,7 @@
 import mock
 import pytest
 
-from datadog_checks.dev import docker_run, get_docker_hostname, get_here
+from datadog_checks.dev import docker_run, get_docker_hostname, get_e2e_discovery_metadata, get_here
 
 HERE = get_here()
 HOST = get_docker_hostname()
@@ -20,7 +20,7 @@ def dd_environment(instance):
 
     with docker_run(compose_file, log_patterns=[r'init - Scylla version \S* initialization completed.']):
         instances = {'instances': [instance]}
-        yield instances
+        yield instances, get_e2e_discovery_metadata()
 
 
 @pytest.fixture(scope="session")

scylla/tests/test_e2e.py

@@ -6,14 +6,22 @@
 
 import pytest
 
+from datadog_checks.dev.docker import CONTAINER_STABILITY_LOG_PATTERNS, assert_all_discovery_candidates_stable
 from datadog_checks.scylla import ScyllaCheck
 
 from .common import FLAKY_METRICS, INSTANCE_DEFAULT_METRICS_V2
 
+# Probing scylla-db's internal storage RPC port (7000) with a plain HTTP GET
+# makes Scylla's own RPC layer log an ERROR ("rpc - client <addr>: server
+# connection dropped: connection is closed") once it rejects the unrecognized
+# protocol magic.
+DISCOVERY_STABILITY_LOG_PATTERNS = tuple(
+    r'error(?!.*rpc - client \S+: server connection dropped: connection is closed)' if pattern == 'error' else pattern
+    for pattern in CONTAINER_STABILITY_LOG_PATTERNS
+)
 
-@pytest.mark.skipif(platform.python_version() < "3", reason='OpenMetrics V2 is only available with Python 3')
-def test_check_ok_omv2(dd_agent_check, instance):
-    aggregator = dd_agent_check(instance, rate=True)
+
+def assert_metrics(aggregator):
     for metric in INSTANCE_DEFAULT_METRICS_V2:
         if metric in FLAKY_METRICS:
             aggregator.assert_metric(metric, at_least=0)
@@ -22,3 +30,23 @@ def test_check_ok_omv2(dd_agent_check, instance):
 
     aggregator.assert_all_metrics_covered()
     aggregator.assert_service_check('scylla.openmetrics.health', ScyllaCheck.OK)
+
+
+@pytest.mark.skipif(platform.python_version() < "3", reason='OpenMetrics V2 is only available with Python 3')
+def test_check_ok_omv2(dd_agent_check, instance):
+    aggregator = dd_agent_check(instance, rate=True)
+    assert_metrics(aggregator)
+
+
+@pytest.mark.skipif(platform.python_version() < "3", reason='OpenMetrics V2 is only available with Python 3')
+@pytest.mark.e2e
+def test_e2e_discovery(dd_agent_check_discovery):
+    aggregator = dd_agent_check_discovery(rate=True)
+    assert_metrics(aggregator)
+
+
+@pytest.mark.e2e
+def test_e2e_discovery_all_candidates(dd_agent_check):
+    assert_all_discovery_candidates_stable(
+        dd_agent_check, ScyllaCheck, compose_service='scylla-db', log_patterns=DISCOVERY_STABILITY_LOG_PATTERNS
+    )