envoy: add connection limits metrics

see also
https://www.envoyproxy.io/docs/envoy/latest/configuration/listeners/network_filters/connection_limit_filter

- Add envoy connection limits metrics
- unit test
- changelog
- Update metadata.csv
- adapt stat_prefix behaviour to 1.29
  see also envoyproxy/envoy@ea71e73

Signed-off-by: William Dauchy <[email protected]>

fix metric map

envoy/changelog.d/16718.added

@@ -0,0 +1 @@
+Add connection limit metrics for envoy

envoy/datadog_checks/envoy/check.py

@@ -17,6 +17,7 @@
     'cluster_name': 'envoy_cluster',
     'envoy_cluster_name': 'envoy_cluster',
     'envoy_local_http_ratelimit_prefix': 'stat_prefix',  # local rate limit
+    'envoy_connection_limit_prefix': 'stat_prefix',  # connection limit
     'envoy_http_conn_manager_prefix': 'stat_prefix',  # tracing
     'envoy_listener_address': 'address',  # listener
     'envoy_virtual_cluster': 'virtual_envoy_cluster',  # vhost
@@ -90,6 +91,16 @@
         'metric_type': 'monotonic_count',
         'new_name': 'listener.downstream_cx.count',
     },
+    r'envoy_connection_limit_(.+)_active_connections$': {
+        'label_name': 'stat_prefix',
+        'metric_type': 'monotonic_count',
+        'new_name': 'connection_limit.active_connections.count',
+    },
+    r'envoy_connection_limit_(.+)_limited_connections$': {
+        'label_name': 'stat_prefix',
+        'metric_type': 'monotonic_count',
+        'new_name': 'connection_limit.limited_connections.count',
+    },
     r'envoy_(.+)_http_local_rate_limit_enabled$': {
         'label_name': 'stat_prefix',
         'metric_type': 'monotonic_count',

envoy/datadog_checks/envoy/metrics.py

@@ -383,6 +383,8 @@
     'envoy_cluster_client_ssl_socket_factory_ssl_context_update_by_sds': 'cluster.client_ssl_socket_factory.ssl_context_update_by_sds',  # noqa: E501
     'envoy_cluster_client_ssl_socket_factory_upstream_context_secrets_not_ready': 'cluster.client_ssl_socket_factory.upstream_context_secrets_not_ready',  # noqa: E501
     'envoy_cluster_client_ssl_socket_factory_downstream_context_secrets_not_ready': 'cluster.client_ssl_socket_factory.downstream_context_secrets_not_ready',  # noqa: E501
+    'envoy_connection_limit_active_connections': 'connection_limit.active_connections',
+    'envoy_connection_limit_limited_connections': 'connection_limit.limited_connections',
 }
 
 # fmt: off
@@ -3933,6 +3935,20 @@
         ),
         'method': 'monotonic_count',
     },
+    'connection_limit.active_connections': {
+        'tags': (
+            ('stat_prefix',),
+            (),
+        ),
+        'method': 'monotonic_count',
+    },
+    'connection_limit.limited_connections': {
+        'tags': (
+            ('stat_prefix',),
+            (),
+        ),
+        'method': 'monotonic_count',
+    },
 }
 # fmt: on
 

envoy/metadata.csv

@@ -149,6 +149,10 @@ envoy.cluster_manager.cluster_updated.count,count,,,,[OpenMetrics V2] Total clus
 envoy.cluster_manager.custer_updated_via_merge.count,count,,,,[OpenMetrics V2],0,envoy,,
 envoy.cluster_manager.update_merge_cancelled.count,count,,,,[OpenMetrics V2] Total merged updates that got cancelled and delivered early,0,envoy,,
 envoy.cluster_manager.update_out_of_merge_window.count,count,,,,[OpenMetrics V2] Total updates which arrived out of a merge window,0,envoy,,
+envoy.connection_limit.active_connections.count,count,,,,[OpenMetrics V2] Number of currently active connections in the scope of this network filter chain,0,envoy,,
+envoy.connection_limit.limited_connections.count,count,,,,[OpenMetrics V2] Total connections that have been rejected due to connection limit exceeded,0,envoy,,
+envoy.connection_limit.active_connections,count,,,,[Legacy] Number of currently active connections in the scope of this network filter chain,0,envoy,,
+envoy.connection_limit.limited_connections,count,,,,[Legacy] Total connections that have been rejected due to connection limit exceeded,0,envoy,,
 envoy.filesystem.flushed_by_timer.count,count,,,,[OpenMetrics V2],0,envoy,,
 envoy.filesystem.reopen_failed.count,count,,,,[OpenMetrics V2],0,envoy,,
 envoy.filesystem.write_buffered.count,count,,,,[OpenMetrics V2],0,envoy,,

envoy/tests/common.py

@@ -376,13 +376,22 @@
     "vhost.vcluster.upstream_rq.count",
 ]
 
+CONNECTION_LIMIT_METRICS = [
+    "connection_limit.active_connections.count",
+    "connection_limit.limited_connections.count",
+]
+
+CONNECTION_LIMIT_STAT_PREFIX_TAG = 'stat_prefix:ingress_http'
+
 LOCAL_RATE_LIMIT_METRICS = [
     "http.local_rate_limit_enabled.count",
     "http.local_rate_limit_enforced.count",
     "http.local_rate_limit_rate_limited.count",
     "http.local_rate_limit_ok.count",
 ]
 
+RATE_LIMIT_STAT_PREFIX_TAG = 'stat_prefix:http_local_rate_limiter'
+
 CLUSTER_AND_LISTENER_SSL_METRICS = [
     "cluster.client_ssl_socket_factory.downstream_context_secrets_not_ready.count",
     "cluster.client_ssl_socket_factory.ssl_context_update_by_sds.count",
@@ -394,8 +403,6 @@
 
 CONNECT_STATE_METRIC = ['control_plane.connected_state']
 
-RATE_LIMIT_STAT_PREFIX_TAG = 'stat_prefix:http_local_rate_limiter'
-
 FLAKY_METRICS = [
     "listener.downstream_cx_active",
     "listener.downstream_cx_destroy.count",
@@ -731,6 +738,8 @@
     "tcp.on_demand_cluster_timeout.count",
     "tcp.upstream_flush.count",
     "tcp.upstream_flush_active",
+    "connection_limit.active_connections.count",
+    "connection_limit.limited_connections.count",
 ]
 
 

envoy/tests/docker/api_v3/front-envoy.yaml

@@ -24,6 +24,12 @@ static_resources:
       socket_address: {address: 0.0.0.0, port_value: 80}
     filter_chains:
     - filters:
+      - name: envoy.filters.network.connection_limit
+        typed_config:
+          "@type": type.googleapis.com/envoy.extensions.filters.network.connection_limit.v3.ConnectionLimit
+          stat_prefix: ingress_http
+          max_connections: 1000
+          delay: 0s
       - name: envoy.http_connection_manager
         typed_config:
           "@type": type.googleapis.com/envoy.extensions.filters.network.http_connection_manager.v3.HttpConnectionManager

envoy/tests/fixtures/legacy/connection_limit.txt

@@ -0,0 +1,2 @@
+connection_limit.ingress_http.active_connections: 0
+connection_limit.ingress_http.limited_connections: 0

envoy/tests/fixtures/legacy/stat_prefix

@@ -8,3 +8,5 @@ cluster.foo.ext_authz.bar.disabled: 6
 cluster.foo.ext_authz.bar.error: 7
 cluster.foo.ext_authz.bar.failure_mode_allowed: 8
 cluster.foo.ext_authz.bar.ok: 9
+connection_limit.ingress_http.active_connections: 0
+connection_limit.ingress_http.limited_connections: 0

envoy/tests/fixtures/openmetrics/openmetrics.txt

@@ -521,6 +521,10 @@ envoy_filesystem_write_buffered{} 7
 envoy_http_downstream_cx_upgrades_total{envoy_http_conn_manager_prefix="admin"} 0
 # TYPE envoy_cluster_manager_cluster_removed counter
 envoy_cluster_manager_cluster_removed{} 0
+# TYPE envoy_connection_limit_ingress_http_active_connections counter
+envoy_connection_limit_ingress_http_active_connections{} 0
+# TYPE envoy_connection_limit_ingress_http_limited_connections counter
+envoy_connection_limit_ingress_http_limited_connections{} 0
 # TYPE envoy_server_debug_assertion_failures counter
 envoy_server_debug_assertion_failures{} 0
 # TYPE envoy_server_worker_3_watchdog_miss counter

envoy/tests/fixtures/openmetrics/openmetrics_1_29.txt

@@ -0,0 +1,4 @@
+# TYPE envoy_connection_limit_active_connections counter
+envoy_connection_limit_active_connections{envoy_connection_limit_prefix="ingress_http"} 0
+# TYPE envoy_connection_limit_limited_connections counter
+envoy_connection_limit_limited_connections{envoy_connection_limit_prefix="ingress_http"} 0

envoy/tests/legacy/common.py

@@ -52,7 +52,14 @@
     "envoy.http_local_rate_limit.ok",
 ]
 
-STAT_PREFIX_TAG = ['stat_prefix:http_local_rate_limiter', 'stat_prefix:foo_buz_112']
+RATE_LIMIT_STAT_PREFIX_TAG = ['stat_prefix:http_local_rate_limiter', 'stat_prefix:foo_buz_112']
+
+CONNECTION_LIMIT_METRICS = [
+    "envoy.connection_limit.active_connections",
+    "envoy.connection_limit.limited_connections",
+]
+
+CONNECTION_LIMIT_STAT_PREFIX_TAG = ['stat_prefix:ingress_http']
 
 RBAC_METRICS = [
     "envoy.http.rbac.allowed",

envoy/tests/legacy/test_unit.py

@@ -12,14 +12,16 @@
 from datadog_checks.envoy.metrics import METRIC_PREFIX, METRICS
 
 from .common import (
+    CONNECTION_LIMIT_METRICS,
+    CONNECTION_LIMIT_STAT_PREFIX_TAG,
     ENVOY_VERSION,
     EXT_METRICS,
     FLAVOR,
     HOST,
     INSTANCES,
     LOCAL_RATE_LIMIT_METRICS,
+    RATE_LIMIT_STAT_PREFIX_TAG,
     RBAC_METRICS,
-    STAT_PREFIX_TAG,
 )
 
 CHECK_NAME = 'envoy'
@@ -289,7 +291,7 @@ def test_stats_prefix_optional_tags(
     standard_tags.append('endpoint:{}'.format(instance["stats_url"]))
     tags_prefix = standard_tags + additional_tags
     c = check(instance)
-    mock_http_response(file_path=fixture_path(fixture_file)).return_value
+    mock_http_response(file_path=fixture_path(fixture_file))
     dd_run_check(c)
 
     # To ensure that this change didn't break the old behavior, both the value and the tags are asserted.
@@ -313,7 +315,20 @@ def test_local_rate_limit_metrics(aggregator, fixture_path, mock_http_response,
 
     for metric in LOCAL_RATE_LIMIT_METRICS:
         aggregator.assert_metric(metric)
-        for tag in STAT_PREFIX_TAG:
+        for tag in RATE_LIMIT_STAT_PREFIX_TAG:
+            aggregator.assert_metric_has_tag(metric, tag, count=1)
+
+    aggregator.assert_metrics_using_metadata(get_metadata_metrics())
+
+
+def test_connection_limit_metrics(aggregator, fixture_path, mock_http_response, check, dd_run_check):
+    instance = INSTANCES['main']
+    c = check(instance)
+
+    mock_http_response(file_path=fixture_path('./legacy/connection_limit.txt'))
+    dd_run_check(c)
+    for metric in CONNECTION_LIMIT_METRICS:
+        for tag in CONNECTION_LIMIT_STAT_PREFIX_TAG:
             aggregator.assert_metric_has_tag(metric, tag, count=1)
 
     aggregator.assert_metrics_using_metadata(get_metadata_metrics())

envoy/tests/legacy/test_utils.py

@@ -74,6 +74,20 @@ def test_wildcard_removal_tree():
                 ()
             ),
             "method": "monotonic_count",
+        },
+        "connection_limit.*.active_connections": {
+            "tags": (
+                ("stat_prefix",),
+                ()
+            ),
+            "method": "monotonic_count",
+        },
+        "connection_limit.*.limited_connections": {
+            "tags": (
+                ("stat_prefix",),
+                ()
+            ),
+            "method": "monotonic_count",
         }
     }
 
@@ -85,6 +99,14 @@ def test_wildcard_removal_tree():
         "http_local_rate_limit.enforced": {
             "tags": (("stat_prefix",), (), ()),
             "method": "monotonic_count",
+        },
+        "connection_limit.active_connections": {
+            "tags": (("stat_prefix",), ()),
+            "method": "monotonic_count",
+        },
+        "connection_limit.limited_connections": {
+            "tags": (("stat_prefix",), ()),
+            "method": "monotonic_count",
         }
     }
     # fmt: on

envoy/tests/test_unit.py

@@ -12,6 +12,8 @@
 from .common import (
     CLUSTER_AND_LISTENER_SSL_METRICS,
     CONNECT_STATE_METRIC,
+    CONNECTION_LIMIT_METRICS,
+    CONNECTION_LIMIT_STAT_PREFIX_TAG,
     DEFAULT_INSTANCE,
     LOCAL_RATE_LIMIT_METRICS,
     MOCKED_PROMETHEUS_METRICS,
@@ -46,7 +48,12 @@ def test_check(aggregator, dd_run_check, check, mock_http_response):
 
     dd_run_check(c)
 
-    for metric in MOCKED_PROMETHEUS_METRICS + LOCAL_RATE_LIMIT_METRICS + CLUSTER_AND_LISTENER_SSL_METRICS:
+    for metric in (
+        MOCKED_PROMETHEUS_METRICS
+        + LOCAL_RATE_LIMIT_METRICS
+        + CONNECTION_LIMIT_METRICS
+        + CLUSTER_AND_LISTENER_SSL_METRICS
+    ):
         aggregator.assert_metric("envoy.{}".format(metric))
 
     for metric in CONNECT_STATE_METRIC:
@@ -130,6 +137,40 @@ def test_local_rate_limit_metrics(aggregator, dd_run_check, check, mock_http_res
     aggregator.assert_metrics_using_metadata(get_metadata_metrics())
 
 
+@requires_py3
+@pytest.mark.parametrize(
+    'fixture_file',
+    [
+        'openmetrics.txt',
+        'openmetrics_1_29.txt',
+    ],
+    ids=[
+        "Envoy < 1.29",
+        "Envoy >= 1.29",
+    ],
+)
+def test_connection_limit_metrics(aggregator, dd_run_check, check, mock_http_response, fixture_file):
+    # Envoy 1.29+ fixed this metric by moving the variable stat_prefix into a label which follows the normal
+    # OpenMetrics convention. However older versions still have the stat_prefix inside the metric name.
+    # https://github.com/envoyproxy/envoy/commit/ea71e737298a03298f478489c181395629a21ce3
+    mock_http_response(file_path=get_fixture_path('./openmetrics/{}'.format(fixture_file)))
+
+    c = check(DEFAULT_INSTANCE)
+
+    dd_run_check(c)
+
+    for metric in CONNECTION_LIMIT_METRICS:
+        aggregator.assert_metric('envoy.{}'.format(metric))
+        aggregator.assert_metric_has_tag('envoy.{}'.format(metric), CONNECTION_LIMIT_STAT_PREFIX_TAG)
+
+    aggregator.assert_service_check(
+        "envoy.openmetrics.health", status=AgentCheck.OK, tags=['endpoint:http://localhost:8001/stats/prometheus']
+    )
+
+    aggregator.assert_no_duplicate_metrics()
+    aggregator.assert_metrics_using_metadata(get_metadata_metrics())
+
+
 @requires_py3
 def test_tags_in_ssl_metrics(aggregator, dd_run_check, check, mock_http_response):
     mock_http_response(file_path=get_fixture_path('./openmetrics/openmetrics_ssl_metrics.txt'))