fix(storage-routing): If item type is unspecified, default to spans (#7350)

Requests for Traces do not have item type specified (which makes sense).
In order to use tracing, a customer is most likely going to be ingesting
spans. Thus, use spans as the default item when unspecified.

We can definitely get more sophisticated with this but this will allow
us to sample GetTraces calls and solve some timeouts on big customers

Author: volokluev

snuba/web/rpc/storage_routing/routing_strategies/outcomes_based.py

@@ -170,12 +170,14 @@ def _get_routing_decision(self, routing_context: RoutingContext) -> RoutingDecis
                     else "normal"
                 ),
             )
-        if (
-            self._is_highest_accuracy_mode(in_msg_meta)
-            or in_msg_meta.trace_item_type not in _ITEM_TYPE_TO_OUTCOME
+        if self._is_highest_accuracy_mode(in_msg_meta) or (
+            # unspecified item type will be assumed as spans when querying
+            # for GetTraces, there is no type specified so we assume spans because
+            # that is necessary for traces anyways
+            # if the type is specified and we don't know its outcome, route to Tier_1
+            in_msg_meta.trace_item_type != TraceItemType.TRACE_ITEM_TYPE_UNSPECIFIED
+            and in_msg_meta.trace_item_type not in _ITEM_TYPE_TO_OUTCOME
         ):
-            if span:
-                span.set_data("tier", routing_decision.tier.name)
             return routing_decision
         # if we're querying a short enough timeframe, don't bother estimating, route to tier 1 and call it a day
         start_ts = in_msg_meta.start_timestamp.seconds

tests/web/rpc/v1/routing_strategies/test_outcomes_based.py

@@ -158,3 +158,26 @@ def test_outcomes_based_routing_highest_accuracy_mode(store_outcomes_data: Any)
     assert routing_decision.tier == Tier.TIER_1
     assert routing_decision.clickhouse_settings == {}
     assert routing_decision.can_run
+
+
+@pytest.mark.clickhouse_db
+@pytest.mark.redis_db
+def test_outcomes_based_routing_defaults_to_spans_for_unspecified_item_type(
+    store_outcomes_data: Any,
+) -> None:
+    strategy = OutcomesBasedRoutingStrategy()
+
+    request = TraceItemTableRequest(meta=_get_request_meta())
+    request.meta.trace_item_type = TraceItemType.TRACE_ITEM_TYPE_UNSPECIFIED
+    state.set_config(
+        "OutcomesBasedRoutingStrategy.max_items_before_downsampling", 50_000
+    )
+    routing_decision = strategy.get_routing_decision(
+        RoutingContext(
+            in_msg=request,
+            timer=Timer("test"),
+        )
+    )
+    assert routing_decision.tier == Tier.TIER_512
+    assert routing_decision.clickhouse_settings == {}
+    assert routing_decision.can_run