Add is_cloud info to telemetry + get consistent customer_uuid's for a… (#3684)

* Add is_cloud info to telemetry + get consistent customer_uuid's for a given tenant

* Address Richard's comments

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

@@ -67,7 +67,7 @@ class Metric(BaseModel):
         }
         task_logger.info(json.dumps(data))
 
-    def emit(self) -> None:
+    def emit(self, tenant_id: str | None) -> None:
         # Convert value to appropriate type
         float_value = (
             float(self.value) if isinstance(self.value, (int, float)) else None
@@ -104,6 +104,7 @@ class Metric(BaseModel):
         optional_telemetry(
             record_type=RecordType.METRIC,
             data=data,
+            tenant_id=tenant_id,
         )
 
 
@@ -432,7 +433,7 @@ def monitor_background_processes(self: Task, *, tenant_id: str | None) -> None:
                 metrics = metric_fn()
                 for metric in metrics:
                     metric.log()
-                    metric.emit()
+                    metric.emit(tenant_id)
                     if metric.key:
                         _mark_metric_as_emitted(redis_std, metric.key)
 

backend/onyx/main.py

@@ -215,7 +215,11 @@ async def lifespan(app: FastAPI) -> AsyncGenerator[None, None]:
     else:
         setup_multitenant_onyx()
 
-    optional_telemetry(record_type=RecordType.VERSION, data={"version": __version__})
+    if not MULTI_TENANT:
+        # don't emit a metric for every pod rollover/restart
+        optional_telemetry(
+            record_type=RecordType.VERSION, data={"version": __version__}
+        )
 
     if AUTH_RATE_LIMITING_ENABLED:
         await setup_auth_limiter()

backend/onyx/onyxbot/slack/utils.py

@@ -590,6 +590,7 @@ def slack_usage_report(
         record_type=RecordType.USAGE,
         data={"action": action},
         user_id=str(onyx_user.id) if onyx_user else "Non-Onyx-Or-No-Auth-User",
+        tenant_id=tenant_id,
     )
 
 

backend/onyx/utils/telemetry.py

@@ -21,6 +21,7 @@ from onyx.utils.variable_functionality import (
 )
 from onyx.utils.variable_functionality import noop_fallback
 from shared_configs.configs import MULTI_TENANT
+from shared_configs.contextvars import CURRENT_TENANT_ID_CONTEXTVAR
 
 _DANSWER_TELEMETRY_ENDPOINT = "https://telemetry.onyx.app/anonymous_telemetry"
 _CACHED_UUID: str | None = None
@@ -36,7 +37,16 @@ class RecordType(str, Enum):
     METRIC = "metric"
 
 
-def get_or_generate_uuid() -> str:
+def _get_or_generate_customer_id_mt(tenant_id: str) -> str:
+    return str(uuid.uuid5(uuid.NAMESPACE_X500, tenant_id))
+
+
+def get_or_generate_uuid(tenant_id: str | None = None) -> str:
+    # TODO: split out the whole "instance UUID" generation logic into a separate
+    # utility function. Telemetry should not be aware at all of how the UUID is
+    # generated/stored.
+    # TODO: handle potential race condition for UUID generation. Doesn't matter for
+    # the telemetry case, but if this is used generally it should be handled.
     global _CACHED_UUID
 
     if _CACHED_UUID is not None:
@@ -53,7 +63,7 @@ def get_or_generate_uuid() -> str:
     return _CACHED_UUID
 
 
-def _get_or_generate_instance_domain() -> str | None:
+def _get_or_generate_instance_domain() -> str | None:  #
     global _CACHED_INSTANCE_DOMAIN
 
     if _CACHED_INSTANCE_DOMAIN is not None:
@@ -76,7 +86,10 @@ def _get_or_generate_instance_domain() -> str | None:
 
 
 def optional_telemetry(
-    record_type: RecordType, data: dict, user_id: str | None = None
+    record_type: RecordType,
+    data: dict,
+    user_id: str | None = None,
+    tenant_id: str | None = None,
 ) -> None:
     if DISABLE_TELEMETRY:
         return
@@ -85,7 +98,13 @@ def optional_telemetry(
 
         def telemetry_logic() -> None:
             try:
-                customer_uuid = get_or_generate_uuid()
+                customer_uuid = (
+                    _get_or_generate_customer_id_mt(
+                        tenant_id or CURRENT_TENANT_ID_CONTEXTVAR.get()
+                    )
+                    if MULTI_TENANT
+                    else get_or_generate_uuid()
+                )
                 payload = {
                     "data": data,
                     "record": record_type,
@@ -93,6 +112,7 @@ def optional_telemetry(
                     # For cases where the User itself is None, a string is provided instead
                     "user_id": user_id,
                     "customer_uuid": customer_uuid,
+                    "is_cloud": MULTI_TENANT,
                 }
                 if ENTERPRISE_EDITION_ENABLED:
                     payload["instance_domain"] = _get_or_generate_instance_domain()