From 9a1376c4c6e33c646a5e2f7fbd1b6e4090128830 Mon Sep 17 00:00:00 2001
From: "copilot-swe-agent[bot]" <198982749+Copilot@users.noreply.github.com>
Date: Thu, 16 Apr 2026 17:41:53 +0000
Subject: [PATCH 01/12] Initial plan


From 6194792895b834d29d23b1f113378971c23f195d Mon Sep 17 00:00:00 2001
From: "copilot-swe-agent[bot]" <198982749+Copilot@users.noreply.github.com>
Date: Thu, 16 Apr 2026 17:51:16 +0000
Subject: [PATCH 02/12] chore: outline plan for MCP gateway health check retry
 fix

Agent-Logs-Url: https://github.com/github/gh-aw/sessions/c923f8eb-a610-4433-b2c6-cfae71b2811d

Co-authored-by: pelikhan <4175913+pelikhan@users.noreply.github.com>
---
 .github/mcp.json | 8 ++++++++
 1 file changed, 8 insertions(+)
 create mode 100644 .github/mcp.json

diff --git a/.github/mcp.json b/.github/mcp.json
new file mode 100644
index 00000000000..9ca83b55e03
--- /dev/null
+++ b/.github/mcp.json
@@ -0,0 +1,8 @@
+{
+  "mcpServers": {
+    "github-agentic-workflows": {
+      "command": "gh",
+      "args": ["aw", "mcp-server"]
+    }
+  }
+}

From 5fd98d19c3ffe3b7af4df53538eaf7df9605d0cb Mon Sep 17 00:00:00 2001
From: "copilot-swe-agent[bot]" <198982749+Copilot@users.noreply.github.com>
Date: Thu, 16 Apr 2026 17:56:55 +0000
Subject: [PATCH 03/12] fix: add MCP gateway health-check retry backoff for
 startup delays

Agent-Logs-Url: https://github.com/github/gh-aw/sessions/c923f8eb-a610-4433-b2c6-cfae71b2811d

Co-authored-by: pelikhan <4175913+pelikhan@users.noreply.github.com>
---
 .github/mcp.json                           |  8 --------
 actions/setup/js/start_mcp_gateway.cjs     | 11 ++++++++---
 actions/setup/sh/start_mcp_gateway.sh      |  9 +++++----
 actions/setup/sh/start_mcp_gateway_test.sh |  7 +++++++
 4 files changed, 20 insertions(+), 15 deletions(-)
 delete mode 100644 .github/mcp.json

diff --git a/.github/mcp.json b/.github/mcp.json
deleted file mode 100644
index 9ca83b55e03..00000000000
--- a/.github/mcp.json
+++ /dev/null
@@ -1,8 +0,0 @@
-{
-  "mcpServers": {
-    "github-agentic-workflows": {
-      "command": "gh",
-      "args": ["aw", "mcp-server"]
-    }
-  }
-}
diff --git a/actions/setup/js/start_mcp_gateway.cjs b/actions/setup/js/start_mcp_gateway.cjs
index 57f2e0b5603..adefe7ad8ca 100644
--- a/actions/setup/js/start_mcp_gateway.cjs
+++ b/actions/setup/js/start_mcp_gateway.cjs
@@ -370,16 +370,20 @@ async function main() {
 
   core.info(`Health endpoint: ${healthUrl}`);
   core.info(`(Note: MCP_GATEWAY_DOMAIN is '${gatewayDomain}' for container access)`);
-  core.info("Retrying up to 120 times with 1s delay (120s total timeout)");
+  core.info("Retrying up to 120 times with exponential backoff (250ms to 1s, ~120s total timeout)");
   core.info("");
 
   const maxRetries = 120;
+  const initialRetryDelayMs = 250;
+  const maxRetryDelayMs = 1000;
   let httpCode = 0;
   let healthBody = "";
   let succeeded = false;
+  let attemptsMade = 0;
 
   core.info("=== Health Check Progress ===");
   for (let attempt = 1; attempt <= maxRetries; attempt++) {
+    attemptsMade = attempt;
     const elapsedSec = Math.floor((nowMs() - healthCheckStart) / 1000);
     if (attempt % 10 === 1 || attempt === 1) {
       core.info(`Attempt ${attempt}/${maxRetries} (${elapsedSec}s elapsed)...`);
@@ -399,14 +403,15 @@ async function main() {
     }
 
     if (attempt < maxRetries) {
-      await sleep(1000);
+      const retryDelayMs = Math.min(initialRetryDelayMs * 2 ** (attempt - 1), maxRetryDelayMs);
+      await sleep(retryDelayMs);
     }
   }
   core.info("=== End Health Check Progress ===");
   core.info("");
 
   core.info(`Final HTTP code: ${httpCode}`);
-  core.info(`Total attempts: ${maxRetries}`);
+  core.info(`Total attempts: ${attemptsMade}`);
   if (healthBody) {
     core.info(`Health response body: ${healthBody}`);
   } else {
diff --git a/actions/setup/sh/start_mcp_gateway.sh b/actions/setup/sh/start_mcp_gateway.sh
index 824bff5a9a7..b232d587424 100755
--- a/actions/setup/sh/start_mcp_gateway.sh
+++ b/actions/setup/sh/start_mcp_gateway.sh
@@ -209,7 +209,7 @@ HEALTH_CHECK_START=$(date +%s%3N)
 HEALTH_CHECK_HOST="localhost"
 echo "Health endpoint: http://${HEALTH_CHECK_HOST}:${MCP_GATEWAY_PORT}/health"
 echo "(Note: MCP_GATEWAY_DOMAIN is '${MCP_GATEWAY_DOMAIN}' for container access)"
-echo "Retrying up to 120 times with 1s delay (120s total timeout)"
+echo "Retrying up to 120 times with exponential backoff (250ms to 1s, ~120s total timeout)"
 echo ""
 
 # Check health endpoint using localhost (since we're running on the host)
@@ -219,7 +219,8 @@ echo ""
 set +e
 
 MAX_RETRIES=120
-RETRY_DELAY=1
+INITIAL_RETRY_DELAY=0.25
+MAX_RETRY_DELAY=1
 RETRY_COUNT=0
 HTTP_CODE=""
 HEALTH_RESPONSE=""
@@ -234,7 +235,6 @@ while [ $RETRY_COUNT -lt $MAX_RETRIES ]; do
   ELAPSED_MS=$(($(date +%s%3N) - HEALTH_CHECK_START))
   ELAPSED_SEC=$((ELAPSED_MS / 1000))
   
-  # Show progress every 10 retries or on first attempt
   if [ $((RETRY_COUNT % 10)) -eq 1 ] || [ $RETRY_COUNT -eq 1 ]; then
     echo "Attempt $RETRY_COUNT/$MAX_RETRIES (${ELAPSED_SEC}s elapsed)..."
   fi
@@ -255,7 +255,8 @@ while [ $RETRY_COUNT -lt $MAX_RETRIES ]; do
   
   # If this is not the last attempt, wait before retrying
   if [ $RETRY_COUNT -lt $MAX_RETRIES ]; then
-    sleep $RETRY_DELAY
+    RETRY_DELAY=$(awk -v base="$INITIAL_RETRY_DELAY" -v exp="$((RETRY_COUNT - 1))" -v max="$MAX_RETRY_DELAY" 'BEGIN { delay = base * (2 ^ exp); if (delay > max) delay = max; printf "%.3f", delay }')
+    sleep "$RETRY_DELAY"
   fi
 done
 echo "=== End Health Check Progress ==="
diff --git a/actions/setup/sh/start_mcp_gateway_test.sh b/actions/setup/sh/start_mcp_gateway_test.sh
index e1d4e1ba565..718fd463880 100755
--- a/actions/setup/sh/start_mcp_gateway_test.sh
+++ b/actions/setup/sh/start_mcp_gateway_test.sh
@@ -281,6 +281,13 @@ test_validation_functions_exist() {
   else
     print_result "--network host flag validation missing" "FAIL"
   fi
+
+  # Check for health check retry/backoff logic
+  if grep -q "INITIAL_RETRY_DELAY=0.25" "$SCRIPT_PATH" && grep -q "MAX_RETRY_DELAY=1" "$SCRIPT_PATH"; then
+    print_result "Health check exponential backoff configuration exists" "PASS"
+  else
+    print_result "Health check exponential backoff configuration missing" "FAIL"
+  fi
 }
 
 # Run all tests

From 70aa68eaa380f5a18c13f0ae6237a2495ec13c8e Mon Sep 17 00:00:00 2001
From: "copilot-swe-agent[bot]" <198982749+Copilot@users.noreply.github.com>
Date: Thu, 16 Apr 2026 17:59:38 +0000
Subject: [PATCH 04/12] refactor: cap MCP gateway startup backoff without
 subprocess overhead

Agent-Logs-Url: https://github.com/github/gh-aw/sessions/c923f8eb-a610-4433-b2c6-cfae71b2811d

Co-authored-by: pelikhan <4175913+pelikhan@users.noreply.github.com>
---
 actions/setup/js/start_mcp_gateway.cjs     |  4 +++-
 actions/setup/sh/start_mcp_gateway.sh      | 12 +++++++++---
 actions/setup/sh/start_mcp_gateway_test.sh |  2 +-
 3 files changed, 13 insertions(+), 5 deletions(-)

diff --git a/actions/setup/js/start_mcp_gateway.cjs b/actions/setup/js/start_mcp_gateway.cjs
index adefe7ad8ca..079067c514b 100644
--- a/actions/setup/js/start_mcp_gateway.cjs
+++ b/actions/setup/js/start_mcp_gateway.cjs
@@ -376,6 +376,7 @@ async function main() {
   const maxRetries = 120;
   const initialRetryDelayMs = 250;
   const maxRetryDelayMs = 1000;
+  const maxRetryExponent = Math.floor(Math.log2(maxRetryDelayMs / initialRetryDelayMs));
   let httpCode = 0;
   let healthBody = "";
   let succeeded = false;
@@ -403,7 +404,8 @@ async function main() {
     }
 
     if (attempt < maxRetries) {
-      const retryDelayMs = Math.min(initialRetryDelayMs * 2 ** (attempt - 1), maxRetryDelayMs);
+      const retryExponent = Math.min(attempt - 1, maxRetryExponent);
+      const retryDelayMs = Math.min(initialRetryDelayMs * 2 ** retryExponent, maxRetryDelayMs);
       await sleep(retryDelayMs);
     }
   }
diff --git a/actions/setup/sh/start_mcp_gateway.sh b/actions/setup/sh/start_mcp_gateway.sh
index b232d587424..604428836d7 100755
--- a/actions/setup/sh/start_mcp_gateway.sh
+++ b/actions/setup/sh/start_mcp_gateway.sh
@@ -219,8 +219,6 @@ echo ""
 set +e
 
 MAX_RETRIES=120
-INITIAL_RETRY_DELAY=0.25
-MAX_RETRY_DELAY=1
 RETRY_COUNT=0
 HTTP_CODE=""
 HEALTH_RESPONSE=""
@@ -255,7 +253,15 @@ while [ $RETRY_COUNT -lt $MAX_RETRIES ]; do
   
   # If this is not the last attempt, wait before retrying
   if [ $RETRY_COUNT -lt $MAX_RETRIES ]; then
-    RETRY_DELAY=$(awk -v base="$INITIAL_RETRY_DELAY" -v exp="$((RETRY_COUNT - 1))" -v max="$MAX_RETRY_DELAY" 'BEGIN { delay = base * (2 ^ exp); if (delay > max) delay = max; printf "%.3f", delay }')
+    # Exponential backoff with 1s cap:
+    # attempt 1 -> 0.25s, attempt 2 -> 0.5s, attempt 3+ -> 1s
+    if [ $RETRY_COUNT -eq 1 ]; then
+      RETRY_DELAY="0.25"
+    elif [ $RETRY_COUNT -eq 2 ]; then
+      RETRY_DELAY="0.5"
+    else
+      RETRY_DELAY="1"
+    fi
     sleep "$RETRY_DELAY"
   fi
 done
diff --git a/actions/setup/sh/start_mcp_gateway_test.sh b/actions/setup/sh/start_mcp_gateway_test.sh
index 718fd463880..4cb9f029310 100755
--- a/actions/setup/sh/start_mcp_gateway_test.sh
+++ b/actions/setup/sh/start_mcp_gateway_test.sh
@@ -283,7 +283,7 @@ test_validation_functions_exist() {
   fi
 
   # Check for health check retry/backoff logic
-  if grep -q "INITIAL_RETRY_DELAY=0.25" "$SCRIPT_PATH" && grep -q "MAX_RETRY_DELAY=1" "$SCRIPT_PATH"; then
+  if grep -q "RETRY_COUNT -eq 1" "$SCRIPT_PATH" && grep -q "RETRY_COUNT -eq 2" "$SCRIPT_PATH" && grep -q "RETRY_DELAY=\"1\"" "$SCRIPT_PATH"; then
     print_result "Health check exponential backoff configuration exists" "PASS"
   else
     print_result "Health check exponential backoff configuration missing" "FAIL"

From 93af3fc09cd269a8e2ed32c3276ef5b8cea9655a Mon Sep 17 00:00:00 2001
From: "copilot-swe-agent[bot]" <198982749+Copilot@users.noreply.github.com>
Date: Thu, 16 Apr 2026 18:01:34 +0000
Subject: [PATCH 05/12] test: tighten MCP gateway backoff assertions

Agent-Logs-Url: https://github.com/github/gh-aw/sessions/c923f8eb-a610-4433-b2c6-cfae71b2811d

Co-authored-by: pelikhan <4175913+pelikhan@users.noreply.github.com>
---
 actions/setup/js/start_mcp_gateway.cjs     | 3 ++-
 actions/setup/sh/start_mcp_gateway_test.sh | 6 +++++-
 2 files changed, 7 insertions(+), 2 deletions(-)

diff --git a/actions/setup/js/start_mcp_gateway.cjs b/actions/setup/js/start_mcp_gateway.cjs
index 079067c514b..8540dec1523 100644
--- a/actions/setup/js/start_mcp_gateway.cjs
+++ b/actions/setup/js/start_mcp_gateway.cjs
@@ -376,7 +376,8 @@ async function main() {
   const maxRetries = 120;
   const initialRetryDelayMs = 250;
   const maxRetryDelayMs = 1000;
-  const maxRetryExponent = Math.floor(Math.log2(maxRetryDelayMs / initialRetryDelayMs));
+  // 250ms -> 500ms -> 1000ms, then capped at 1000ms.
+  const maxRetryExponent = 2;
   let httpCode = 0;
   let healthBody = "";
   let succeeded = false;
diff --git a/actions/setup/sh/start_mcp_gateway_test.sh b/actions/setup/sh/start_mcp_gateway_test.sh
index 4cb9f029310..62efb969f00 100755
--- a/actions/setup/sh/start_mcp_gateway_test.sh
+++ b/actions/setup/sh/start_mcp_gateway_test.sh
@@ -283,7 +283,11 @@ test_validation_functions_exist() {
   fi
 
   # Check for health check retry/backoff logic
-  if grep -q "RETRY_COUNT -eq 1" "$SCRIPT_PATH" && grep -q "RETRY_COUNT -eq 2" "$SCRIPT_PATH" && grep -q "RETRY_DELAY=\"1\"" "$SCRIPT_PATH"; then
+  if grep -q "RETRY_COUNT -eq 1" "$SCRIPT_PATH" &&
+    grep -q "RETRY_COUNT -eq 2" "$SCRIPT_PATH" &&
+    grep -q "RETRY_DELAY=\"0.25\"" "$SCRIPT_PATH" &&
+    grep -q "RETRY_DELAY=\"0.5\"" "$SCRIPT_PATH" &&
+    grep -q "RETRY_DELAY=\"1\"" "$SCRIPT_PATH"; then
     print_result "Health check exponential backoff configuration exists" "PASS"
   else
     print_result "Health check exponential backoff configuration missing" "FAIL"

From f5cb3d500e79fc27a4952537600ea371cd706344 Mon Sep 17 00:00:00 2001
From: "copilot-swe-agent[bot]" <198982749+Copilot@users.noreply.github.com>
Date: Thu, 16 Apr 2026 18:03:18 +0000
Subject: [PATCH 06/12] chore: simplify MCP gateway JS backoff calculation

Agent-Logs-Url: https://github.com/github/gh-aw/sessions/c923f8eb-a610-4433-b2c6-cfae71b2811d

Co-authored-by: pelikhan <4175913+pelikhan@users.noreply.github.com>
---
 actions/setup/js/start_mcp_gateway.cjs | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/actions/setup/js/start_mcp_gateway.cjs b/actions/setup/js/start_mcp_gateway.cjs
index 8540dec1523..abec6fbf267 100644
--- a/actions/setup/js/start_mcp_gateway.cjs
+++ b/actions/setup/js/start_mcp_gateway.cjs
@@ -375,7 +375,7 @@ async function main() {
 
   const maxRetries = 120;
   const initialRetryDelayMs = 250;
-  const maxRetryDelayMs = 1000;
+  // delay = initialRetryDelayMs * 2^min(attempt-1, maxRetryExponent)
   // 250ms -> 500ms -> 1000ms, then capped at 1000ms.
   const maxRetryExponent = 2;
   let httpCode = 0;
@@ -406,7 +406,7 @@ async function main() {
 
     if (attempt < maxRetries) {
       const retryExponent = Math.min(attempt - 1, maxRetryExponent);
-      const retryDelayMs = Math.min(initialRetryDelayMs * 2 ** retryExponent, maxRetryDelayMs);
+      const retryDelayMs = initialRetryDelayMs * 2 ** retryExponent;
       await sleep(retryDelayMs);
     }
   }

From 2d880a4cf30f1aa535662a55167ac74ca6e0b11c Mon Sep 17 00:00:00 2001
From: "copilot-swe-agent[bot]" <198982749+Copilot@users.noreply.github.com>
Date: Thu, 16 Apr 2026 18:05:21 +0000
Subject: [PATCH 07/12] docs: clarify backoff formula and strengthen shell
 pattern checks

Agent-Logs-Url: https://github.com/github/gh-aw/sessions/c923f8eb-a610-4433-b2c6-cfae71b2811d

Co-authored-by: pelikhan <4175913+pelikhan@users.noreply.github.com>
---
 actions/setup/js/start_mcp_gateway.cjs     | 2 +-
 actions/setup/sh/start_mcp_gateway_test.sh | 5 ++++-
 2 files changed, 5 insertions(+), 2 deletions(-)

diff --git a/actions/setup/js/start_mcp_gateway.cjs b/actions/setup/js/start_mcp_gateway.cjs
index abec6fbf267..ff5bde4811d 100644
--- a/actions/setup/js/start_mcp_gateway.cjs
+++ b/actions/setup/js/start_mcp_gateway.cjs
@@ -375,7 +375,7 @@ async function main() {
 
   const maxRetries = 120;
   const initialRetryDelayMs = 250;
-  // delay = initialRetryDelayMs * 2^min(attempt-1, maxRetryExponent)
+  // delay = initialRetryDelayMs * 2**min(attempt-1, maxRetryExponent)
   // 250ms -> 500ms -> 1000ms, then capped at 1000ms.
   const maxRetryExponent = 2;
   let httpCode = 0;
diff --git a/actions/setup/sh/start_mcp_gateway_test.sh b/actions/setup/sh/start_mcp_gateway_test.sh
index 62efb969f00..cdee99198b8 100755
--- a/actions/setup/sh/start_mcp_gateway_test.sh
+++ b/actions/setup/sh/start_mcp_gateway_test.sh
@@ -285,9 +285,12 @@ test_validation_functions_exist() {
   # Check for health check retry/backoff logic
   if grep -q "RETRY_COUNT -eq 1" "$SCRIPT_PATH" &&
     grep -q "RETRY_COUNT -eq 2" "$SCRIPT_PATH" &&
+    grep -q "elif \[ \$RETRY_COUNT -eq 2 \]" "$SCRIPT_PATH" &&
+    grep -q "else" "$SCRIPT_PATH" &&
     grep -q "RETRY_DELAY=\"0.25\"" "$SCRIPT_PATH" &&
     grep -q "RETRY_DELAY=\"0.5\"" "$SCRIPT_PATH" &&
-    grep -q "RETRY_DELAY=\"1\"" "$SCRIPT_PATH"; then
+    grep -q "RETRY_DELAY=\"1\"" "$SCRIPT_PATH" &&
+    grep -q "sleep \"\$RETRY_DELAY\"" "$SCRIPT_PATH"; then
     print_result "Health check exponential backoff configuration exists" "PASS"
   else
     print_result "Health check exponential backoff configuration missing" "FAIL"

From d8744749fa248dd591751656e613767deed4dda5 Mon Sep 17 00:00:00 2001
From: "copilot-swe-agent[bot]" <198982749+Copilot@users.noreply.github.com>
Date: Thu, 16 Apr 2026 18:06:47 +0000
Subject: [PATCH 08/12] test: assert documented third backoff tier

Agent-Logs-Url: https://github.com/github/gh-aw/sessions/c923f8eb-a610-4433-b2c6-cfae71b2811d

Co-authored-by: pelikhan <4175913+pelikhan@users.noreply.github.com>
---
 actions/setup/sh/start_mcp_gateway_test.sh | 1 +
 1 file changed, 1 insertion(+)

diff --git a/actions/setup/sh/start_mcp_gateway_test.sh b/actions/setup/sh/start_mcp_gateway_test.sh
index cdee99198b8..2712646be53 100755
--- a/actions/setup/sh/start_mcp_gateway_test.sh
+++ b/actions/setup/sh/start_mcp_gateway_test.sh
@@ -290,6 +290,7 @@ test_validation_functions_exist() {
     grep -q "RETRY_DELAY=\"0.25\"" "$SCRIPT_PATH" &&
     grep -q "RETRY_DELAY=\"0.5\"" "$SCRIPT_PATH" &&
     grep -q "RETRY_DELAY=\"1\"" "$SCRIPT_PATH" &&
+    grep -q "attempt 3+ -> 1s" "$SCRIPT_PATH" &&
     grep -q "sleep \"\$RETRY_DELAY\"" "$SCRIPT_PATH"; then
     print_result "Health check exponential backoff configuration exists" "PASS"
   else

From 0bb933393578dd6d7f0c0a4f4b174ce17c7db181 Mon Sep 17 00:00:00 2001
From: "copilot-swe-agent[bot]" <198982749+Copilot@users.noreply.github.com>
Date: Thu, 16 Apr 2026 18:36:33 +0000
Subject: [PATCH 09/12] chore: plan follow-up for review comment

Agent-Logs-Url: https://github.com/github/gh-aw/sessions/0e241816-4acc-4ea2-98b5-d6db16911e8a

Co-authored-by: pelikhan <4175913+pelikhan@users.noreply.github.com>
---
 .github/mcp.json | 11 +++++++++++
 1 file changed, 11 insertions(+)
 create mode 100644 .github/mcp.json

diff --git a/.github/mcp.json b/.github/mcp.json
new file mode 100644
index 00000000000..b953af2639e
--- /dev/null
+++ b/.github/mcp.json
@@ -0,0 +1,11 @@
+{
+  "mcpServers": {
+    "github-agentic-workflows": {
+      "command": "gh",
+      "args": [
+        "aw",
+        "mcp-server"
+      ]
+    }
+  }
+}
\ No newline at end of file

From 7682cd261aa7f1a4f2582799584683b8bb10f9aa Mon Sep 17 00:00:00 2001
From: "copilot-swe-agent[bot]" <198982749+Copilot@users.noreply.github.com>
Date: Thu, 16 Apr 2026 18:40:00 +0000
Subject: [PATCH 10/12] refactor: use shared withRetry helper in MCP gateway
 health polling

Agent-Logs-Url: https://github.com/github/gh-aw/sessions/0e241816-4acc-4ea2-98b5-d6db16911e8a

Co-authored-by: pelikhan <4175913+pelikhan@users.noreply.github.com>
---
 .github/mcp.json                       | 11 -----
 actions/setup/js/start_mcp_gateway.cjs | 61 ++++++++++++++------------
 2 files changed, 33 insertions(+), 39 deletions(-)
 delete mode 100644 .github/mcp.json

diff --git a/.github/mcp.json b/.github/mcp.json
deleted file mode 100644
index b953af2639e..00000000000
--- a/.github/mcp.json
+++ /dev/null
@@ -1,11 +0,0 @@
-{
-  "mcpServers": {
-    "github-agentic-workflows": {
-      "command": "gh",
-      "args": [
-        "aw",
-        "mcp-server"
-      ]
-    }
-  }
-}
\ No newline at end of file
diff --git a/actions/setup/js/start_mcp_gateway.cjs b/actions/setup/js/start_mcp_gateway.cjs
index ff5bde4811d..c81c950b36d 100644
--- a/actions/setup/js/start_mcp_gateway.cjs
+++ b/actions/setup/js/start_mcp_gateway.cjs
@@ -31,6 +31,7 @@ const { spawn, execSync } = require("child_process");
 const fs = require("fs");
 const http = require("http");
 const path = require("path");
+const { withRetry } = require("./error_recovery.cjs");
 
 // ---------------------------------------------------------------------------
 // Timing helpers
@@ -373,43 +374,47 @@ async function main() {
   core.info("Retrying up to 120 times with exponential backoff (250ms to 1s, ~120s total timeout)");
   core.info("");
 
-  const maxRetries = 120;
+  const maxAttempts = 120;
   const initialRetryDelayMs = 250;
-  // delay = initialRetryDelayMs * 2**min(attempt-1, maxRetryExponent)
-  // 250ms -> 500ms -> 1000ms, then capped at 1000ms.
-  const maxRetryExponent = 2;
   let httpCode = 0;
   let healthBody = "";
   let succeeded = false;
   let attemptsMade = 0;
 
   core.info("=== Health Check Progress ===");
-  for (let attempt = 1; attempt <= maxRetries; attempt++) {
-    attemptsMade = attempt;
-    const elapsedSec = Math.floor((nowMs() - healthCheckStart) / 1000);
-    if (attempt % 10 === 1 || attempt === 1) {
-      core.info(`Attempt ${attempt}/${maxRetries} (${elapsedSec}s elapsed)...`);
-    }
-
-    try {
-      const res = await httpGet(healthUrl, 2000);
-      httpCode = res.statusCode;
-      healthBody = res.body;
-      if (httpCode === 200 && healthBody) {
-        core.info(`✓ Health check succeeded on attempt ${attempt} (${elapsedSec}s elapsed)`);
-        succeeded = true;
-        break;
-      }
-    } catch {
-      // Connection refused / timeout – retry
-    }
+  try {
+    await withRetry(
+      async () => {
+        attemptsMade += 1;
+        const elapsedSec = Math.floor((nowMs() - healthCheckStart) / 1000);
+        if (attemptsMade % 10 === 1 || attemptsMade === 1) {
+          core.info(`Attempt ${attemptsMade}/${maxAttempts} (${elapsedSec}s elapsed)...`);
+        }
 
-    if (attempt < maxRetries) {
-      const retryExponent = Math.min(attempt - 1, maxRetryExponent);
-      const retryDelayMs = initialRetryDelayMs * 2 ** retryExponent;
-      await sleep(retryDelayMs);
-    }
+        const res = await httpGet(healthUrl, 2000);
+        httpCode = res.statusCode;
+        healthBody = res.body;
+        if (httpCode === 200 && healthBody) {
+          core.info(`✓ Health check succeeded on attempt ${attemptsMade} (${elapsedSec}s elapsed)`);
+          succeeded = true;
+          return;
+        }
+        throw new Error(`Health endpoint not ready (HTTP ${httpCode || 0})`);
+      },
+      {
+        maxRetries: maxAttempts - 1,
+        initialDelayMs: initialRetryDelayMs,
+        maxDelayMs: 1000,
+        backoffMultiplier: 2,
+        jitterMs: 0,
+        shouldRetry: () => true,
+      },
+      "MCP gateway health check"
+    );
+  } catch {
+    // Retry exhaustion is handled below using existing diagnostics.
   }
+
   core.info("=== End Health Check Progress ===");
   core.info("");
 

From ebf5c12be84edbadfe155435963acb35a6f27eb0 Mon Sep 17 00:00:00 2001
From: "copilot-swe-agent[bot]" <198982749+Copilot@users.noreply.github.com>
Date: Thu, 16 Apr 2026 18:41:46 +0000
Subject: [PATCH 11/12] chore: clarify withRetry attempt semantics in gateway
 health check

Agent-Logs-Url: https://github.com/github/gh-aw/sessions/0e241816-4acc-4ea2-98b5-d6db16911e8a

Co-authored-by: pelikhan <4175913+pelikhan@users.noreply.github.com>
---
 actions/setup/js/start_mcp_gateway.cjs | 9 ++++++---
 1 file changed, 6 insertions(+), 3 deletions(-)

diff --git a/actions/setup/js/start_mcp_gateway.cjs b/actions/setup/js/start_mcp_gateway.cjs
index c81c950b36d..3ba71812074 100644
--- a/actions/setup/js/start_mcp_gateway.cjs
+++ b/actions/setup/js/start_mcp_gateway.cjs
@@ -374,7 +374,9 @@ async function main() {
   core.info("Retrying up to 120 times with exponential backoff (250ms to 1s, ~120s total timeout)");
   core.info("");
 
-  const maxAttempts = 120;
+  const maxTotalAttempts = 120;
+  // withRetry's maxRetries excludes the initial attempt.
+  const maxRetryCount = maxTotalAttempts - 1;
   const initialRetryDelayMs = 250;
   let httpCode = 0;
   let healthBody = "";
@@ -385,10 +387,11 @@ async function main() {
   try {
     await withRetry(
       async () => {
+        // Counts total health-check attempts, including the final successful attempt.
         attemptsMade += 1;
         const elapsedSec = Math.floor((nowMs() - healthCheckStart) / 1000);
         if (attemptsMade % 10 === 1 || attemptsMade === 1) {
-          core.info(`Attempt ${attemptsMade}/${maxAttempts} (${elapsedSec}s elapsed)...`);
+          core.info(`Attempt ${attemptsMade}/${maxTotalAttempts} (${elapsedSec}s elapsed)...`);
         }
 
         const res = await httpGet(healthUrl, 2000);
@@ -402,7 +405,7 @@ async function main() {
         throw new Error(`Health endpoint not ready (HTTP ${httpCode || 0})`);
       },
       {
-        maxRetries: maxAttempts - 1,
+        maxRetries: maxRetryCount,
         initialDelayMs: initialRetryDelayMs,
         maxDelayMs: 1000,
         backoffMultiplier: 2,

From 7b4184f05e42ec0c2caa604544a56dde2dad1cae Mon Sep 17 00:00:00 2001
From: "copilot-swe-agent[bot]" <198982749+Copilot@users.noreply.github.com>
Date: Thu, 16 Apr 2026 18:43:31 +0000
Subject: [PATCH 12/12] docs: clarify intentional retry policy for gateway
 health polling

Agent-Logs-Url: https://github.com/github/gh-aw/sessions/0e241816-4acc-4ea2-98b5-d6db16911e8a

Co-authored-by: pelikhan <4175913+pelikhan@users.noreply.github.com>
---
 actions/setup/js/start_mcp_gateway.cjs | 1 +
 1 file changed, 1 insertion(+)

diff --git a/actions/setup/js/start_mcp_gateway.cjs b/actions/setup/js/start_mcp_gateway.cjs
index 3ba71812074..bfca155fb2f 100644
--- a/actions/setup/js/start_mcp_gateway.cjs
+++ b/actions/setup/js/start_mcp_gateway.cjs
@@ -410,6 +410,7 @@ async function main() {
         maxDelayMs: 1000,
         backoffMultiplier: 2,
         jitterMs: 0,
+        // Preserve previous loop behavior: retry any health-check failure until attempts are exhausted.
         shouldRetry: () => true,
       },
       "MCP gateway health check"