Update go.mod and go.sum

Signed-off-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>

admin/dashboard.html

@@ -1277,15 +1277,34 @@
             stateEl.classList.remove('loading');
 
             let badgeClass = 'badge-info';
-            if (data.state === 'closed') badgeClass = 'badge-success';
-            else if (data.state === 'open') badgeClass = 'badge-danger';
-            else if (data.state === 'half-open') badgeClass = 'badge-warning';
+            let stateText = data.state || 'unknown';
 
-            stateEl.innerHTML = `<span class="badge ${badgeClass}">${data.state || 'Unknown'}</span>`;
+            // For cluster mode, determine state from instance counts
+            if (data.instances_open !== undefined) {
+                // Cluster mode data
+                if (data.instances_open > 0) {
+                    stateText = `${data.instances_open} open`;
+                    badgeClass = 'badge-danger';
+                } else if (data.instances_halfopen > 0) {
+                    stateText = `${data.instances_halfopen} half-open`;
+                    badgeClass = 'badge-warning';
+                } else if (data.instances_closed > 0) {
+                    stateText = `${data.instances_closed} closed`;
+                    badgeClass = 'badge-success';
+                }
+            } else {
+                // Single instance mode
+                if (data.state === 'closed') badgeClass = 'badge-success';
+                else if (data.state === 'open') badgeClass = 'badge-danger';
+                else if (data.state === 'half-open') badgeClass = 'badge-warning';
+            }
+
+            stateEl.innerHTML = `<span class="badge ${badgeClass}">${stateText}</span>`;
 
             document.getElementById('cb-enabled').textContent = data.enabled ? 'Yes' : 'No';
 
             if (data.counts) {
+                // Single instance mode with detailed counts
                 document.getElementById('cb-total-requests').textContent =
                     (data.counts.requests || 0).toLocaleString();
                 document.getElementById('cb-total-successes').textContent =
@@ -1296,6 +1315,14 @@
                     (data.counts.consecutive_successes || 0).toLocaleString();
                 document.getElementById('cb-consecutive-failures').textContent =
                     (data.counts.consecutive_failures || 0).toLocaleString();
+            } else if (data.instances_open !== undefined) {
+                // Cluster mode - show instance distribution instead
+                const total = (data.instances_open || 0) + (data.instances_closed || 0) + (data.instances_halfopen || 0);
+                document.getElementById('cb-total-requests').textContent = total + ' instances';
+                document.getElementById('cb-total-successes').textContent = (data.instances_closed || 0).toLocaleString();
+                document.getElementById('cb-total-failures').textContent = (data.instances_open || 0).toLocaleString();
+                document.getElementById('cb-consecutive-successes').textContent = '--';
+                document.getElementById('cb-consecutive-failures').textContent = '--';
             }
 
             if (data.config) {
@@ -1307,23 +1334,31 @@
         function updateCoalescing(data) {
             document.getElementById('coalescing-rate').textContent =
                 (data.backend_savings_pct || 0).toFixed(1) + '%';
-            document.getElementById('coalescing-total').textContent =
-                (data.total_requests || 0).toLocaleString();
-            document.getElementById('coalescing-primary').textContent =
-                (data.primary_requests || 0).toLocaleString();
-            document.getElementById('coalescing-coalesced').textContent =
-                (data.coalesced_requests || 0).toLocaleString();
+
+            // Handle both single instance (total_requests) and cluster mode (total_coalesced + total_primary)
+            const totalRequests = data.total_requests ||
+                ((data.total_coalesced_requests || 0) + (data.total_primary_requests || 0));
+            document.getElementById('coalescing-total').textContent = totalRequests.toLocaleString();
+
+            // Handle both single instance and cluster mode field names
+            const primaryRequests = data.primary_requests || data.total_primary_requests || 0;
+            document.getElementById('coalescing-primary').textContent = primaryRequests.toLocaleString();
+
+            const coalescedRequests = data.coalesced_requests || data.total_coalesced_requests || 0;
+            document.getElementById('coalescing-coalesced').textContent = coalescedRequests.toLocaleString();
+
             document.getElementById('coalescing-savings').textContent =
                 (data.backend_savings_pct || 0).toFixed(1) + '%';
         }
 
         function updateRetryBudget(data) {
-            document.getElementById('retry-tokens').textContent =
-                data.current_tokens || '--';
-            document.getElementById('retry-current-tokens').textContent =
-                data.current_tokens || '--';
-            document.getElementById('retry-max-tokens').textContent =
-                data.max_tokens || '--';
+            // Use explicit undefined check to handle 0 values correctly
+            const currentTokens = data.current_tokens !== undefined ? data.current_tokens : '--';
+            const maxTokens = data.max_tokens !== undefined ? data.max_tokens : '--';
+
+            document.getElementById('retry-tokens').textContent = currentTokens;
+            document.getElementById('retry-current-tokens').textContent = currentTokens;
+            document.getElementById('retry-max-tokens').textContent = maxTokens;
             document.getElementById('retry-total').textContent =
                 (data.total_attempts || 0).toLocaleString();
             document.getElementById('retry-denied').textContent =
@@ -1333,13 +1368,17 @@
         }
 
         function updateWebSocket(data) {
-            document.getElementById('ws-connections').textContent =
-                data.active_connections || 0;
+            // Handle both single instance (active_connections) and cluster mode (total_connections)
+            const connections = data.active_connections !== undefined ? data.active_connections :
+                               (data.total_connections !== undefined ? data.total_connections : 0);
+            document.getElementById('ws-connections').textContent = connections;
         }
 
         function updateConnections(data) {
-            document.getElementById('pool-connections').textContent =
-                data.active_connections || 0;
+            // Handle both single instance (active_connections) and cluster mode (total_active)
+            const connections = data.active_connections !== undefined ? data.active_connections :
+                               (data.total_active !== undefined ? data.total_active : 0);
+            document.getElementById('pool-connections').textContent = connections;
         }
 
         async function resetCoalescing() {

go.mod

@@ -20,7 +20,7 @@ require (
 	github.com/lukaszraczylo/ask v0.0.0-20240916204100-6e9ef53a62d9
 	github.com/lukaszraczylo/go-ratecounter v0.1.12
 	github.com/lukaszraczylo/go-simple-graphql v1.2.89
-	github.com/redis/go-redis/v9 v9.17.1
+	github.com/redis/go-redis/v9 v9.17.2
 	github.com/sony/gobreaker v1.0.0
 	github.com/stretchr/testify v1.11.1
 	github.com/valyala/fasthttp v1.68.0
@@ -47,7 +47,7 @@ require (
 	github.com/jackc/pgpassfile v1.0.0 // indirect
 	github.com/jackc/pgservicefile v0.0.0-20240606120523-5a60cdf6a761 // indirect
 	github.com/jackc/puddle/v2 v2.2.2 // indirect
-	github.com/klauspost/compress v1.18.1 // indirect
+	github.com/klauspost/compress v1.18.2 // indirect
 	github.com/mattn/go-colorable v0.1.14 // indirect
 	github.com/mattn/go-isatty v0.0.20 // indirect
 	github.com/mattn/go-runewidth v0.0.19 // indirect

go.sum

@@ -64,8 +64,8 @@ github.com/jackc/pgx/v5 v5.7.6 h1:rWQc5FwZSPX58r1OQmkuaNicxdmExaEz5A2DO2hUuTk=
 github.com/jackc/pgx/v5 v5.7.6/go.mod h1:aruU7o91Tc2q2cFp5h4uP3f6ztExVpyVv88Xl/8Vl8M=
 github.com/jackc/puddle/v2 v2.2.2 h1:PR8nw+E/1w0GLuRFSmiioY6UooMp6KJv0/61nB7icHo=
 github.com/jackc/puddle/v2 v2.2.2/go.mod h1:vriiEXHvEE654aYKXXjOvZM39qJ0q+azkZFrfEOc3H4=
-github.com/klauspost/compress v1.18.1 h1:bcSGx7UbpBqMChDtsF28Lw6v/G94LPrrbMbdC3JH2co=
-github.com/klauspost/compress v1.18.1/go.mod h1:ZQFFVG+MdnR0P+l6wpXgIL4NTtwiKIdBnrBd8Nrxr+0=
+github.com/klauspost/compress v1.18.2 h1:iiPHWW0YrcFgpBYhsA6D1+fqHssJscY/Tm/y2Uqnapk=
+github.com/klauspost/compress v1.18.2/go.mod h1:R0h/fSBs8DE4ENlcrlib3PsXS61voFxhIs2DeRhCvJ4=
 github.com/kr/pretty v0.3.1 h1:flRD4NNwYAUpkphVc1HcthR4KEIFJ65n8Mw5qdRn3LE=
 github.com/kr/pretty v0.3.1/go.mod h1:hoEshYVHaxMs3cyo3Yncou5ZscifuDolrwPKZanG3xk=
 github.com/kr/text v0.2.0 h1:5Nx0Ya0ZqY2ygV366QzturHI13Jq95ApcVaJBhpS+AY=
@@ -84,8 +84,8 @@ github.com/mattn/go-runewidth v0.0.19 h1:v++JhqYnZuu5jSKrk9RbgF5v4CGUjqRfBm05byF
 github.com/mattn/go-runewidth v0.0.19/go.mod h1:XBkDxAl56ILZc9knddidhrOlY5R/pDhgLpndooCuJAs=
 github.com/pmezard/go-difflib v1.0.0 h1:4DBwDE0NGyQoBHbLQYPwSUPoCMWR5BEzIk/f1lZbAQM=
 github.com/pmezard/go-difflib v1.0.0/go.mod h1:iKH77koFhYxTK1pcRnkKkqfTogsbg7gZNVY4sRDYZ/4=
-github.com/redis/go-redis/v9 v9.17.1 h1:7tl732FjYPRT9H9aNfyTwKg9iTETjWjGKEJ2t/5iWTs=
-github.com/redis/go-redis/v9 v9.17.1/go.mod h1:u410H11HMLoB+TP67dz8rL9s6QW2j76l0//kSOd3370=
+github.com/redis/go-redis/v9 v9.17.2 h1:P2EGsA4qVIM3Pp+aPocCJ7DguDHhqrXNhVcEp4ViluI=
+github.com/redis/go-redis/v9 v9.17.2/go.mod h1:u410H11HMLoB+TP67dz8rL9s6QW2j76l0//kSOd3370=
 github.com/rogpeppe/go-internal v1.14.1 h1:UQB4HGPB6osV0SQTLymcB4TgvyWu6ZyliaW0tI/otEQ=
 github.com/rogpeppe/go-internal v1.14.1/go.mod h1:MaRKkUm5W0goXpeCfT7UZI6fk/L7L7so1lCWt35ZSgc=
 github.com/savsgio/gotils v0.0.0-20250924091648-bce9a52d7761 h1:McifyVxygw1d67y6vxUqls2D46J8W9nrki9c8c0eVvE=

metrics_aggregator.go

@@ -519,7 +519,10 @@ func (ma *MetricsAggregator) aggregateStats(instances []InstanceMetrics) map[str
 		totalRetryAllowed  int64
 		totalRetryDenied   int64
 		totalRetryAttempts int64
+		totalCurrentTokens int64
+		totalMaxTokens     int64
 		retryBudgetEnabled = false
+		retryTokensPerSec  float64 // Use max tokens_per_sec from any instance
 
 		// Circuit breaker stats
 		cbOpenCount           int
@@ -645,6 +648,17 @@ func (ma *MetricsAggregator) aggregateStats(instances []InstanceMetrics) map[str
 			if attempts, ok := instance.RetryBudget["total_attempts"].(float64); ok {
 				totalRetryAttempts += int64(attempts)
 			}
+			if currentTokens, ok := instance.RetryBudget["current_tokens"].(float64); ok {
+				totalCurrentTokens += int64(currentTokens)
+			}
+			if maxTokens, ok := instance.RetryBudget["max_tokens"].(float64); ok {
+				totalMaxTokens += int64(maxTokens)
+			}
+			if tokensPerSec, ok := instance.RetryBudget["tokens_per_sec"].(float64); ok {
+				if tokensPerSec > retryTokensPerSec {
+					retryTokensPerSec = tokensPerSec
+				}
+			}
 		}
 
 		// Aggregate circuit breaker stats
@@ -748,6 +762,9 @@ func (ma *MetricsAggregator) aggregateStats(instances []InstanceMetrics) map[str
 			"denied_retries":  totalRetryDenied,
 			"total_attempts":  totalRetryAttempts,
 			"denial_rate_pct": retryDenialRate,
+			"current_tokens":  totalCurrentTokens,
+			"max_tokens":      totalMaxTokens,
+			"tokens_per_sec":  retryTokensPerSec,
 		},
 		"circuit_breaker": map[string]interface{}{
 			"enabled":            circuitBreakerEnabled,