fix(core-backend): improve traces

Author: Kriys94

packages/core-backend/CHANGELOG.md

@@ -7,6 +7,24 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [Unreleased]
 
+### Changed
+
+- Improve WebSocket connection lifecycle tracing in `BackendWebSocketService` ([#7101](https://github.com/MetaMask/core/pull/7101))
+  - WebSocket connection duration is now properly reflected in trace span duration instead of only in custom data
+  - Trace all disconnections (both manual and unexpected) to provide complete connection lifecycle visibility in traces
+  - Omit `connectionDuration_ms` from disconnection traces when connection never established (onClose without onOpen)
+- Update `BackendWebSocketService` default exponential backoff options for reconnection ([#7101](https://github.com/MetaMask/core/pull/7101))
+  - Increase default `reconnectDelay` from 500 milliseconds to 10 seconds
+  - Increase default `maxReconnectDelay` from 30 seconds to 60 seconds
+- Simplify WebSocket disconnection code in `BackendWebSocketService` ([#7101](https://github.com/MetaMask/core/pull/7101))
+  - Centralize all disconnection logic in `ws.onclose` handler for single source of truth
+  - Centralize all state changes within `#establishConnection` method - state transitions only occur in `onopen` (CONNECTING → CONNECTED) and `onclose` (any state → DISCONNECTED)
+  - Add `MANUAL_DISCONNECT_CODE` (4999) and `MANUAL_DISCONNECT_REASON` constants to distinguish manual from unexpected disconnects
+
+### Removed
+
+- Remove `BackendWebSocketService Channel Message` trace as it provided no useful performance insights ([#7101](https://github.com/MetaMask/core/pull/7101))
+
 ## [4.0.0]
 
 ### Changed

packages/core-backend/README.md

@@ -311,7 +311,7 @@ sequenceDiagram
 #### Key Flow Characteristics
 
 1. **Initial Setup**: BackendWebSocketService establishes connection, then AccountActivityService subscribes to selected account. Backend automatically sends a system notification with all chains that are currently up. AccountActivityService tracks these chains internally and notifies TokenBalancesController, which increases polling interval to 5 min
-2. **Chain Status Tracking**: AccountActivityService maintains an internal set of chains that are 'up' based on system notifications. On disconnect/error, it marks all tracked chains as 'down' before clearing the set
+2. **Chain Status Tracking**: AccountActivityService maintains an internal set of chains that are 'up' based on system notifications. On disconnect, it marks all tracked chains as 'down' before clearing the set
 3. **System Notifications**: Backend automatically sends chain status updates (up/down) upon subscription and when status changes. AccountActivityService forwards these to TokenBalancesController, which adjusts polling intervals (up: 5min, down: 30s + immediate fetch)
 4. **User Account Changes**: When users switch accounts, AccountActivityService unsubscribes from old account and subscribes to new account. Backend sends fresh system notification with current chain status for the new account
 5. **Connection Resilience**: On reconnection, AccountActivityService resubscribes to selected account and receives fresh chain status via system notification. Automatic reconnection with exponential backoff

packages/core-backend/src/AccountActivityService.test.ts

@@ -686,11 +686,11 @@ describe('AccountActivityService', () => {
             timestamp: 1760344704595,
           });
 
-          // Publish WebSocket ERROR state event - should flush tracked chains as down
+          // Publish WebSocket DISCONNECTED state event - should flush tracked chains as down
           rootMessenger.publish(
             'BackendWebSocketService:connectionStateChanged',
             {
-              state: WebSocketState.ERROR,
+              state: WebSocketState.DISCONNECTED,
               url: 'ws://test',
               reconnectAttempts: 2,
               timeout: 10000,
@@ -701,7 +701,7 @@ describe('AccountActivityService', () => {
           );
           await completeAsyncOperations(100);
 
-          // Verify that the ERROR state triggered the status change for tracked chains
+          // Verify that the DISCONNECTED state triggered the status change for tracked chains
           expect(statusChangedEventListener).toHaveBeenCalledWith({
             chainIds: ['eip155:1', 'eip155:137', 'eip155:56'],
             status: 'down',
@@ -720,11 +720,11 @@ describe('AccountActivityService', () => {
 
           mocks.getSelectedAccount.mockReturnValue(null);
 
-          // Publish WebSocket ERROR state event without any tracked chains
+          // Publish WebSocket DISCONNECTED state event without any tracked chains
           rootMessenger.publish(
             'BackendWebSocketService:connectionStateChanged',
             {
-              state: WebSocketState.ERROR,
+              state: WebSocketState.DISCONNECTED,
               url: 'ws://test',
               reconnectAttempts: 2,
               timeout: 10000,

packages/core-backend/src/AccountActivityService.ts

@@ -451,6 +451,15 @@ export class AccountActivityService {
       status: data.status,
       timestamp,
     });
+
+    log(
+      `WebSocket status change - Published tracked chains as ${data.status}`,
+      {
+        count: data.chainIds.length,
+        chains: data.chainIds,
+        status: data.status,
+      },
+    );
   }
 
   /**
@@ -467,11 +476,8 @@ export class AccountActivityService {
       // WebSocket connected - resubscribe to selected account
       // The system notification will automatically provide the list of chains that are up
       await this.#subscribeToSelectedAccount();
-    } else if (
-      state === WebSocketState.DISCONNECTED ||
-      state === WebSocketState.ERROR
-    ) {
-      // On disconnect/error, flush all tracked chains as down
+    } else if (state === WebSocketState.DISCONNECTED) {
+      // On disconnect, flush all tracked chains as down
       const chainsToMarkDown = Array.from(this.#chainsUp);
 
       if (chainsToMarkDown.length > 0) {
@@ -481,13 +487,10 @@ export class AccountActivityService {
           timestamp: Date.now(),
         });
 
-        log(
-          'WebSocket error/disconnection - Published tracked chains as down',
-          {
-            count: chainsToMarkDown.length,
-            chains: chainsToMarkDown,
-          },
-        );
+        log('WebSocket disconnection - Published tracked chains as down', {
+          count: chainsToMarkDown.length,
+          chains: chainsToMarkDown,
+        });
 
         // Clear the tracking set since all chains are now down
         this.#chainsUp.clear();