From c5c372e23ad93d04d8bd836073ea2ae821f2771f Mon Sep 17 00:00:00 2001
From: Piotr Sarnacki <drogus@gmail.com>
Date: Mon, 10 Nov 2025 11:58:16 +0100
Subject: [PATCH] Change spacetime_worker_wasm_instance_errors_total metric's
 labels

There are two purpose for introducing the changes in this commit:

1. Decrease the cardinality of the metric, as caller id and connection
   id combination would result in way too many combinations
2. Add a database_identity label, which is much more useful for the
   metric
---
 .../src/host/wasm_common/module_host_actor.rs | 35 ++++---------------
 crates/core/src/worker_metrics/mod.rs         |  4 +--
 2 files changed, 8 insertions(+), 31 deletions(-)

diff --git a/crates/core/src/host/wasm_common/module_host_actor.rs b/crates/core/src/host/wasm_common/module_host_actor.rs
index 30d5819a303..8b0a2b20756 100644
--- a/crates/core/src/host/wasm_common/module_host_actor.rs
+++ b/crates/core/src/host/wasm_common/module_host_actor.rs
@@ -575,12 +575,7 @@ impl InstanceCommon {
 
                 WORKER_METRICS
                     .wasm_instance_errors
-                    .with_label_values(
-                        &caller_identity,
-                        &self.info.module_hash,
-                        &caller_connection_id,
-                        procedure_name,
-                    )
+                    .with_label_values(&self.info.database_identity, &self.info.module_hash, procedure_name)
                     .inc();
 
                 // TODO(procedure-energy):
@@ -686,12 +681,7 @@ impl InstanceCommon {
             Err(ExecutionError::Recoverable(err) | ExecutionError::Trap(err)) => {
                 inst.log_traceback("reducer", reducer_name, &err);
 
-                self.handle_outer_error(
-                    &result.stats.energy,
-                    &caller_identity,
-                    &Some(caller_connection_id),
-                    reducer_name,
-                )
+                self.handle_outer_error(&result.stats.energy, reducer_name)
             }
             Err(ExecutionError::User(err)) => {
                 log_reducer_error(inst.replica_ctx(), timestamp, reducer_name, &err);
@@ -766,21 +756,10 @@ impl InstanceCommon {
         (res, trapped)
     }
 
-    fn handle_outer_error(
-        &mut self,
-        energy: &EnergyStats,
-        caller_identity: &Identity,
-        caller_connection_id: &Option<ConnectionId>,
-        reducer_name: &str,
-    ) -> EventStatus {
+    fn handle_outer_error(&mut self, energy: &EnergyStats, reducer_name: &str) -> EventStatus {
         WORKER_METRICS
             .wasm_instance_errors
-            .with_label_values(
-                caller_identity,
-                &self.info.module_hash,
-                &caller_connection_id.unwrap_or(ConnectionId::ZERO),
-                reducer_name,
-            )
+            .with_label_values(&self.info.database_identity, &self.info.module_hash, reducer_name)
             .inc();
 
         if energy.remaining.get() == 0 {
@@ -903,14 +882,12 @@ impl InstanceCommon {
         let outcome = match (result.call_result, sender) {
             (Err(ExecutionError::Recoverable(err) | ExecutionError::Trap(err)), _) => {
                 inst.log_traceback("view", &view_name, &err);
-                self.handle_outer_error(&result.stats.energy, &caller, &None, &view_name)
-                    .into()
+                self.handle_outer_error(&result.stats.energy, &view_name).into()
             }
             // TODO: maybe do something else with user errors?
             (Err(ExecutionError::User(err)), _) => {
                 inst.log_traceback("view", &view_name, &anyhow::anyhow!(err));
-                self.handle_outer_error(&result.stats.energy, &caller, &None, &view_name)
-                    .into()
+                self.handle_outer_error(&result.stats.energy, &view_name).into()
             }
             // Materialize anonymous view
             (Ok(bytes), None) => {
diff --git a/crates/core/src/worker_metrics/mod.rs b/crates/core/src/worker_metrics/mod.rs
index cf1ec661a23..0f20a12d632 100644
--- a/crates/core/src/worker_metrics/mod.rs
+++ b/crates/core/src/worker_metrics/mod.rs
@@ -3,7 +3,7 @@ use crate::messages::control_db::HostType;
 use once_cell::sync::Lazy;
 use prometheus::{GaugeVec, HistogramVec, IntCounterVec, IntGaugeVec};
 use spacetimedb_datastore::execution_context::WorkloadType;
-use spacetimedb_lib::{ConnectionId, Identity};
+use spacetimedb_lib::Identity;
 use spacetimedb_metrics::metrics_group;
 use spacetimedb_sats::memory_usage::MemoryUsage;
 use spacetimedb_table::page_pool::PagePool;
@@ -243,7 +243,7 @@ metrics_group!(
 
         #[name = spacetime_worker_wasm_instance_errors_total]
         #[help = "The number of fatal WASM instance errors, such as reducer panics."]
-        #[labels(caller_identity: Identity, module_hash: Hash, caller_connection_id: ConnectionId, reducer_symbol: str)]
+        #[labels(database_identity: Identity, module_hash: Hash, reducer_symbol: str)]
         pub wasm_instance_errors: IntCounterVec,
 
         #[name = spacetime_worker_wasm_memory_bytes]