feat(metrics): add connector_name label to push-based connector metrics (#10237)

backend/Dockerfile

@@ -12,7 +12,7 @@ founders@onyx.app for more information. Please visit https://github.com/onyx-dot
 ARG ENABLE_CRAFT=false
 
 # DO_NOT_TRACK is used to disable telemetry for Unstructured
-ENV ONYX_RUNNING_IN_DOCKER="true" \
+ENV DANSWER_RUNNING_IN_DOCKER="true" \
     DO_NOT_TRACK="true" \
     PLAYWRIGHT_BROWSERS_PATH="/app/.cache/ms-playwright"
 

backend/Dockerfile.model_server

@@ -1,7 +1,7 @@
 # Base stage with dependencies
 FROM python:3.11-slim-bookworm@sha256:9c6f90801e6b68e772b7c0ca74260cbf7af9f320acec894e26fccdaccfbe3b47 AS base
 
-ENV ONYX_RUNNING_IN_DOCKER="true" \
+ENV DANSWER_RUNNING_IN_DOCKER="true" \
     HF_HOME=/app/.cache/huggingface
 
 COPY --from=ghcr.io/astral-sh/uv:0.9.9 /uv /uvx /bin/
@@ -50,6 +50,10 @@ COPY ./onyx/utils/logger.py /app/onyx/utils/logger.py
 COPY ./onyx/utils/middleware.py /app/onyx/utils/middleware.py
 COPY ./onyx/utils/tenant.py /app/onyx/utils/tenant.py
 
+# Sentry configuration (used when SENTRY_DSN is set)
+COPY ./onyx/configs/__init__.py /app/onyx/configs/__init__.py
+COPY ./onyx/configs/sentry.py /app/onyx/configs/sentry.py
+
 # Place to fetch version information
 COPY ./onyx/__init__.py /app/onyx/__init__.py
 

backend/onyx/background/celery/memory_monitoring.py

@@ -5,8 +5,8 @@ from logging.handlers import RotatingFileHandler
 
 import psutil
 
+from onyx.utils.logger import is_running_in_container
 from onyx.utils.logger import setup_logger
-from onyx.utils.platform import is_running_in_container
 
 # Regular application logger
 logger = setup_logger()

backend/onyx/background/celery/tasks/docprocessing/tasks.py

@@ -537,10 +537,12 @@ def check_indexing_completion(
             )
 
         source = cc_pair.connector.source.value
+        connector_name = cc_pair.connector.name or f"cc_pair_{cc_pair.id}"
         on_index_attempt_status_change(
             tenant_id=tenant_id,
             source=source,
             cc_pair_id=cc_pair.id,
+            connector_name=connector_name,
             status=attempt.status.value,
         )
 
@@ -568,6 +570,7 @@ def check_indexing_completion(
                 tenant_id=tenant_id,
                 source=source,
                 cc_pair_id=cc_pair.id,
+                connector_name=connector_name,
                 docs_indexed=attempt.new_docs_indexed or 0,
                 success_timestamp=attempt.time_updated.timestamp(),
             )
@@ -595,6 +598,7 @@ def check_indexing_completion(
                     tenant_id=tenant_id,
                     source=source,
                     cc_pair_id=cc_pair.id,
+                    connector_name=connector_name,
                     in_error=False,
                 )
 
@@ -920,10 +924,14 @@ def check_for_indexing(self: Task, *, tenant_id: str) -> int | None:
                         cc_pair_id=cc_pair_id,
                         in_repeated_error_state=True,
                     )
+                    error_connector_name = (
+                        cc_pair.connector.name or f"cc_pair_{cc_pair.id}"
+                    )
                     on_connector_error_state_change(
                         tenant_id=tenant_id,
                         source=cc_pair.connector.source.value,
                         cc_pair_id=cc_pair_id,
+                        connector_name=error_connector_name,
                         in_error=True,
                     )
 

backend/onyx/background/celery/tasks/monitoring/tasks.py

@@ -42,7 +42,7 @@ from onyx.db.models import UserGroup
 from onyx.db.search_settings import get_active_search_settings_list
 from onyx.redis.redis_pool import get_redis_client
 from onyx.redis.redis_pool import redis_lock_dump
-from onyx.utils.platform import is_running_in_container
+from onyx.utils.logger import is_running_in_container
 from onyx.utils.telemetry import optional_telemetry
 from onyx.utils.telemetry import RecordType
 from shared_configs.configs import MULTI_TENANT

backend/onyx/background/celery/tasks/shared/tasks.py

@@ -248,6 +248,7 @@ def document_by_cc_pair_cleanup_task(
                         ),
                     )
                     mark_document_as_modified(document_id, db_session)
+                    db_session.commit()
                 completion_status = (
                     OnyxCeleryTaskCompletionStatus.NON_RETRYABLE_EXCEPTION
                 )

backend/onyx/background/indexing/run_docfetching.py

@@ -273,6 +273,7 @@ def run_docfetching_entrypoint(
             tenant_id=tenant_id,
             source=attempt.connector_credential_pair.connector.source.value,
             cc_pair_id=connector_credential_pair_id,
+            connector_name=connector_name or f"cc_pair_{connector_credential_pair_id}",
             status="in_progress",
         )
 

backend/onyx/server/metrics/connector_health_metrics.py

@@ -7,7 +7,12 @@ to avoid disrupting the caller's business logic.
 Gauge metrics (error state, last success timestamp) are per-process.
 With multiple worker pods, use max() aggregation in PromQL to get the
 correct value across instances, e.g.:
-    max by (cc_pair_id) (onyx_connector_in_error_state)
+    max by (cc_pair_id, connector_name) (onyx_connector_in_error_state)
+
+Unlike the per-task counters in indexing_task_metrics.py, these metrics
+include connector_name because their cardinality is bounded by the number
+of connectors (one series per connector), not by the number of task
+executions.
 """
 
 from prometheus_client import Counter
@@ -17,12 +22,14 @@ from onyx.utils.logger import setup_logger
 
 logger = setup_logger()
 
+_CONNECTOR_LABELS = ["tenant_id", "source", "cc_pair_id", "connector_name"]
+
 # --- Index attempt lifecycle ---
 
 INDEX_ATTEMPT_STATUS = Counter(
     "onyx_index_attempt_transitions_total",
     "Index attempt status transitions",
-    ["tenant_id", "source", "cc_pair_id", "status"],
+    [*_CONNECTOR_LABELS, "status"],
 )
 
 # --- Connector health ---
@@ -30,25 +37,25 @@ INDEX_ATTEMPT_STATUS = Counter(
 CONNECTOR_IN_ERROR_STATE = Gauge(
     "onyx_connector_in_error_state",
     "Whether the connector is in a repeated error state (1=yes, 0=no)",
-    ["tenant_id", "source", "cc_pair_id"],
+    _CONNECTOR_LABELS,
 )
 
 CONNECTOR_LAST_SUCCESS_TIMESTAMP = Gauge(
     "onyx_connector_last_success_timestamp_seconds",
     "Unix timestamp of last successful indexing for this connector",
-    ["tenant_id", "source", "cc_pair_id"],
+    _CONNECTOR_LABELS,
 )
 
 CONNECTOR_DOCS_INDEXED = Counter(
     "onyx_connector_docs_indexed_total",
     "Total documents indexed per connector (monotonic)",
-    ["tenant_id", "source", "cc_pair_id"],
+    _CONNECTOR_LABELS,
 )
 
 CONNECTOR_INDEXING_ERRORS = Counter(
     "onyx_connector_indexing_errors_total",
     "Total failed index attempts per connector (monotonic)",
-    ["tenant_id", "source", "cc_pair_id"],
+    _CONNECTOR_LABELS,
 )
 
 
@@ -56,6 +63,7 @@ def on_index_attempt_status_change(
     tenant_id: str,
     source: str,
     cc_pair_id: int,
+    connector_name: str,
     status: str,
 ) -> None:
     """Called on any index attempt status transition."""
@@ -64,6 +72,7 @@ def on_index_attempt_status_change(
             "tenant_id": tenant_id,
             "source": source,
             "cc_pair_id": str(cc_pair_id),
+            "connector_name": connector_name,
         }
         INDEX_ATTEMPT_STATUS.labels(**labels, status=status).inc()
         if status == "failed":
@@ -76,6 +85,7 @@ def on_connector_error_state_change(
     tenant_id: str,
     source: str,
     cc_pair_id: int,
+    connector_name: str,
     in_error: bool,
 ) -> None:
     """Called when a connector's in_repeated_error_state changes."""
@@ -84,6 +94,7 @@ def on_connector_error_state_change(
             tenant_id=tenant_id,
             source=source,
             cc_pair_id=str(cc_pair_id),
+            connector_name=connector_name,
         ).set(1.0 if in_error else 0.0)
     except Exception:
         logger.debug("Failed to record connector error state metric", exc_info=True)
@@ -93,6 +104,7 @@ def on_connector_indexing_success(
     tenant_id: str,
     source: str,
     cc_pair_id: int,
+    connector_name: str,
     docs_indexed: int,
     success_timestamp: float,
 ) -> None:
@@ -102,6 +114,7 @@ def on_connector_indexing_success(
             "tenant_id": tenant_id,
             "source": source,
             "cc_pair_id": str(cc_pair_id),
+            "connector_name": connector_name,
         }
         CONNECTOR_LAST_SUCCESS_TIMESTAMP.labels(**labels).set(success_timestamp)
         if docs_indexed > 0:

backend/onyx/server/settings/api.py

@@ -34,7 +34,6 @@ from onyx.server.settings.models import UserSettings
 from onyx.server.settings.store import load_settings
 from onyx.server.settings.store import store_settings
 from onyx.utils.logger import setup_logger
-from onyx.utils.platform import is_running_in_container
 from onyx.utils.variable_functionality import (
     fetch_versioned_implementation_with_fallback,
 )
@@ -112,7 +111,6 @@ def fetch_settings(
             if DISABLE_VECTOR_DB
             else DEFAULT_FILE_TOKEN_COUNT_THRESHOLD_K_VECTOR_DB
         ),
-        is_containerized=is_running_in_container(),
     )
 
 

backend/onyx/server/settings/models.py

@@ -133,7 +133,3 @@ class UserSettings(Settings):
             else DEFAULT_FILE_TOKEN_COUNT_THRESHOLD_K_VECTOR_DB
         )
     )
-    # True when the backend is running inside a container (Docker/Podman).
-    # The frontend uses this to default local-service URLs (e.g. Ollama,
-    # LM Studio) to host.docker.internal instead of localhost.
-    is_containerized: bool = False

backend/onyx/utils/logger.py

@@ -5,7 +5,6 @@ from collections.abc import MutableMapping
 from logging.handlers import RotatingFileHandler
 from typing import Any
 
-from onyx.utils.platform import is_running_in_container
 from onyx.utils.tenant import get_tenant_id_short_string
 from shared_configs.configs import DEV_LOGGING_ENABLED
 from shared_configs.configs import LOG_FILE_NAME
@@ -170,6 +169,13 @@ def get_standard_formatter() -> ColoredFormatter:
     )
 
 
+DANSWER_DOCKER_ENV_STR = "DANSWER_RUNNING_IN_DOCKER"
+
+
+def is_running_in_container() -> bool:
+    return os.getenv(DANSWER_DOCKER_ENV_STR) == "true"
+
+
 def setup_logger(
     name: str = __name__,
     log_level: int = get_log_level_from_str(),

backend/onyx/utils/platform.py

@@ -1,32 +0,0 @@
-import logging
-import os
-
-logger = logging.getLogger(__name__)
-
-_ONYX_DOCKER_ENV_STR = "ONYX_RUNNING_IN_DOCKER"
-_DANSWER_DOCKER_ENV_STR = "DANSWER_RUNNING_IN_DOCKER"
-
-
-def _resolve_container_flag() -> bool:
-    onyx_val = os.getenv(_ONYX_DOCKER_ENV_STR)
-    if onyx_val is not None:
-        return onyx_val.lower() == "true"
-
-    danswer_val = os.getenv(_DANSWER_DOCKER_ENV_STR)
-    if danswer_val is not None:
-        logger.warning(
-            "%s is deprecated and will be ignored in a future release. "
-            "Use %s instead.",
-            _DANSWER_DOCKER_ENV_STR,
-            _ONYX_DOCKER_ENV_STR,
-        )
-        return danswer_val.lower() == "true"
-
-    return False
-
-
-_IS_RUNNING_IN_CONTAINER: bool = _resolve_container_flag()
-
-
-def is_running_in_container() -> bool:
-    return _IS_RUNNING_IN_CONTAINER

docs/METRICS.md

@@ -217,19 +217,19 @@ Enriches docfetching and docprocessing tasks with connector-level labels. Silent
 | `onyx_indexing_task_completed_total`  | Counter   | `task_name`, `source`, `tenant_id`, `cc_pair_id`, `outcome` | Indexing tasks completed per connector   |
 | `onyx_indexing_task_duration_seconds` | Histogram | `task_name`, `source`, `tenant_id`                          | Indexing task duration by connector type |
 
-`connector_name` is intentionally excluded from these push-based counters to avoid unbounded cardinality (it's a free-form user string).
+`connector_name` is intentionally excluded from these per-task counters to avoid unbounded cardinality (it's a free-form user string).
 
 ### Connector Health Metrics (`onyx.server.metrics.connector_health_metrics`)
 
-Push-based metrics emitted by docfetching and docprocessing workers at the point where connector state changes occur. Scales to any number of tenants (no schema iteration).
+Push-based metrics emitted by docfetching and docprocessing workers at the point where connector state changes occur. Scales to any number of tenants (no schema iteration). Unlike the per-task counters above, these include `connector_name` because their cardinality is bounded by the number of connectors (one series per connector), not by the number of task executions.
 
-| Metric                                          | Type    | Labels                                        | Description                                                   |
-| ----------------------------------------------- | ------- | --------------------------------------------- | ------------------------------------------------------------- |
-| `onyx_index_attempt_transitions_total`          | Counter | `tenant_id`, `source`, `cc_pair_id`, `status` | Index attempt status transitions (in_progress, success, etc.) |
-| `onyx_connector_in_error_state`                 | Gauge   | `tenant_id`, `source`, `cc_pair_id`           | Whether connector is in repeated error state (1=yes, 0=no)    |
-| `onyx_connector_last_success_timestamp_seconds` | Gauge   | `tenant_id`, `source`, `cc_pair_id`           | Unix timestamp of last successful indexing                    |
-| `onyx_connector_docs_indexed_total`             | Counter | `tenant_id`, `source`, `cc_pair_id`           | Total documents indexed per connector (monotonic)             |
-| `onyx_connector_indexing_errors_total`          | Counter | `tenant_id`, `source`, `cc_pair_id`           | Total failed index attempts per connector (monotonic)         |
+| Metric                                          | Type    | Labels                                                          | Description                                                   |
+| ----------------------------------------------- | ------- | --------------------------------------------------------------- | ------------------------------------------------------------- |
+| `onyx_index_attempt_transitions_total`          | Counter | `tenant_id`, `source`, `cc_pair_id`, `connector_name`, `status` | Index attempt status transitions (in_progress, success, etc.) |
+| `onyx_connector_in_error_state`                 | Gauge   | `tenant_id`, `source`, `cc_pair_id`, `connector_name`           | Whether connector is in repeated error state (1=yes, 0=no)    |
+| `onyx_connector_last_success_timestamp_seconds` | Gauge   | `tenant_id`, `source`, `cc_pair_id`, `connector_name`           | Unix timestamp of last successful indexing                    |
+| `onyx_connector_docs_indexed_total`             | Counter | `tenant_id`, `source`, `cc_pair_id`, `connector_name`           | Total documents indexed per connector (monotonic)             |
+| `onyx_connector_indexing_errors_total`          | Counter | `tenant_id`, `source`, `cc_pair_id`, `connector_name`           | Total failed index attempts per connector (monotonic)         |
 
 ### Pull-Based Collectors (`onyx.server.metrics.indexing_pipeline`)
 

web/src/interfaces/settings.ts

@@ -76,10 +76,6 @@ export interface Settings {
   // Factory defaults for the restore button.
   default_user_file_max_upload_size_mb?: number;
   default_file_token_count_threshold_k?: number;
-
-  // True when the backend runs inside a container (Docker/Podman).
-  // Used to default local-service URLs to host.docker.internal.
-  is_containerized?: boolean;
 }
 
 export enum NotificationType {

web/src/sections/modals/llmConfig/LMStudioModal.tsx

@@ -27,7 +27,8 @@ import {
 import { fetchModels } from "@/lib/llmConfig/svc";
 import { toast } from "@/hooks/useToast";
 import { refreshLlmProviderCaches } from "@/lib/llmConfig/cache";
-import { useSettingsContext } from "@/providers/SettingsProvider";
+
+const DEFAULT_API_BASE = "http://localhost:1234";
 
 interface LMStudioModalValues extends BaseLLMModalValues {
   api_base: string;
@@ -115,10 +116,6 @@ export default function LMStudioModal({
 }: LLMProviderFormProps) {
   const isOnboarding = variant === "onboarding";
   const { mutate } = useSWRConfig();
-  const { settings } = useSettingsContext();
-  const defaultApiBase = settings.is_containerized
-    ? "http://host.docker.internal:1234"
-    : "http://localhost:1234";
 
   const onClose = () => onOpenChange?.(false);
 
@@ -128,7 +125,7 @@ export default function LMStudioModal({
       LLMProviderName.LM_STUDIO,
       existingLlmProvider
     ),
-    api_base: existingLlmProvider?.api_base ?? defaultApiBase,
+    api_base: existingLlmProvider?.api_base ?? DEFAULT_API_BASE,
     custom_config: {
       LM_STUDIO_API_KEY: existingLlmProvider?.custom_config?.LM_STUDIO_API_KEY,
     },

web/src/sections/modals/llmConfig/OllamaModal.tsx

@@ -31,7 +31,8 @@ import { Card } from "@opal/components";
 import { toast } from "@/hooks/useToast";
 import { refreshLlmProviderCaches } from "@/lib/llmConfig/cache";
 import InputTypeInField from "@/refresh-components/form/InputTypeInField";
-import { useSettingsContext } from "@/providers/SettingsProvider";
+
+const DEFAULT_API_BASE = "http://127.0.0.1:11434";
 const CLOUD_API_BASE = "https://ollama.com";
 
 enum Tab {
@@ -162,10 +163,6 @@ export default function OllamaModal({
 }: LLMProviderFormProps) {
   const isOnboarding = variant === "onboarding";
   const { mutate } = useSWRConfig();
-  const { settings } = useSettingsContext();
-  const defaultApiBase = settings.is_containerized
-    ? "http://host.docker.internal:11434"
-    : "http://127.0.0.1:11434";
   const apiKey = existingLlmProvider?.custom_config?.OLLAMA_API_KEY;
   const defaultTab =
     existingLlmProvider && !!apiKey ? Tab.TAB_CLOUD : Tab.TAB_SELF_HOSTED;
@@ -179,7 +176,7 @@ export default function OllamaModal({
       LLMProviderName.OLLAMA_CHAT,
       existingLlmProvider
     ),
-    api_base: existingLlmProvider?.api_base ?? defaultApiBase,
+    api_base: existingLlmProvider?.api_base ?? DEFAULT_API_BASE,
     custom_config: {
       OLLAMA_API_KEY: apiKey,
     },