From abc4725a7d2903c29cb1fa447e4f290b7b0fefa6 Mon Sep 17 00:00:00 2001
From: Ayush-Patel-56 <ayushpatel2731@gmail.com>
Date: Thu, 19 Feb 2026 20:56:19 +0530
Subject: [PATCH] fix: support serverless logs and robust timeout in diagnose
 scripts

Signed-off-by: Ayush-Patel-56 <ayushpatel2731@gmail.com>
---
 tools/diagnose-fluid-alluxio.sh | 47 ++++++++++++++++++++++++++++++-
 tools/diagnose-fluid-goosefs.sh | 49 +++++++++++++++++++++++++++++++--
 tools/diagnose-fluid-jindo.sh   | 47 ++++++++++++++++++++++++++++++-
 tools/diagnose-fluid-juicefs.sh | 47 ++++++++++++++++++++++++++++++-
 4 files changed, 185 insertions(+), 5 deletions(-)

diff --git a/tools/diagnose-fluid-alluxio.sh b/tools/diagnose-fluid-alluxio.sh
index 4fe88d013a7..910b7dcc04b 100644
--- a/tools/diagnose-fluid-alluxio.sh
+++ b/tools/diagnose-fluid-alluxio.sh
@@ -21,7 +21,16 @@ print_usage() {
 run() {
   echo
   echo "-----------------run $*------------------"
-  timeout 10s "$@"
+  if command -v timeout >/dev/null 2>&1; then
+    timeout 10s "$@"
+  elif command -v gtimeout >/dev/null 2>&1; then
+    gtimeout 10s "$@"
+  elif command -v perl >/dev/null 2>&1; then
+    # Use Perl to enforce timeout on systems without GNU coreutils (like standard macOS)
+    perl -e 'alarm shift; exec @ARGV' 10 "$@"
+  else
+    "$@"
+  fi
   if [ $? != 0 ]; then
     echo "failed to collect info: $*"
   fi
@@ -57,6 +66,41 @@ runtime_pod_logs() {
   core_component "${runtime_namespace}" "alluxio-fuse" "role=alluxio-fuse" "release=${runtime_name}"
 }
 
+serverless_pod_logs() {
+  # Check if kubectl is available
+  if ! command -v kubectl >/dev/null 2>&1; then
+    echo "kubectl not found, skipping serverless pod logs collection"
+    return
+  fi
+
+  local namespace="${runtime_namespace}"
+  # More specific selector: injected and matching the specific dataset/runtime
+  local label_selector="serverless.fluid.io/inject=true,fluid.io/dataset=${runtime_namespace}.${runtime_name}"
+
+  # Get all pods with the serverless inject label
+  local pods
+  pods=$(kubectl get po -n "${namespace}" -l "${label_selector}" -o jsonpath='{.items[*].metadata.name}' 2>/dev/null)
+  if [[ $? -ne 0 ]]; then
+    echo "failed to get serverless pods in namespace ${namespace}"
+    return
+  fi
+
+  if [[ -n "$pods" ]]; then
+    mkdir -p "$diagnose_dir/pods-${namespace}-serverless"
+    for po in ${pods}; do
+      # Find all containers containing fluid-fuse (covers init-fluid-fuse and fluid-fuse)
+      local containers
+      containers=$(kubectl get po "${po}" -n "${namespace}" -o jsonpath='{.spec.initContainers[*].name} {.spec.containers[*].name}' 2>/dev/null | tr ' ' '\n' | grep 'fluid-fuse')
+      if [[ -n "$containers" ]]; then
+        for container in ${containers}; do
+          kubectl logs "${po}" -c "${container}" -n "${namespace}" &>"$diagnose_dir/pods-${namespace}-serverless/${po}-${container}.log" 2>&1
+        done
+      fi
+    done
+  fi
+  return
+}
+
 core_component() {
   # namespace container selectors...
   local namespace="$1"
@@ -105,6 +149,7 @@ pd_collect() {
   pod_status "${fluid_namespace}"
   pod_status "${runtime_namespace}"
   runtime_pod_logs
+  serverless_pod_logs
   fluid_pod_logs
   kubectl_resource
   archive
diff --git a/tools/diagnose-fluid-goosefs.sh b/tools/diagnose-fluid-goosefs.sh
index 1c16bdcfbe1..712606a28a2 100755
--- a/tools/diagnose-fluid-goosefs.sh
+++ b/tools/diagnose-fluid-goosefs.sh
@@ -19,7 +19,16 @@ print_usage() {
 run() {
   echo
   echo "-----------------run $*------------------"
-  timeout 10s "$@"
+  if command -v timeout >/dev/null 2>&1; then
+    timeout 10s "$@"
+  elif command -v gtimeout >/dev/null 2>&1; then
+    gtimeout 10s "$@"
+  elif command -v perl >/dev/null 2>&1; then
+    # Use Perl to enforce timeout on systems without GNU coreutils (like standard macOS)
+    perl -e 'alarm shift; exec @ARGV' 10 "$@"
+  else
+    "$@"
+  fi
   if [ $? != 0 ]; then
     echo "failed to collect info: $*"
   fi
@@ -50,6 +59,41 @@ runtime_pod_logs() {
   core_component "${runtime_namespace}" "goosefs-fuse" "role=goosefs-fuse" "release=${runtime_name}"
 }
 
+serverless_pod_logs() {
+  # Check if kubectl is available
+  if ! command -v kubectl >/dev/null 2>&1; then
+    echo "kubectl not found, skipping serverless pod logs collection"
+    return
+  fi
+
+  local namespace="${runtime_namespace}"
+  # More specific selector: injected and matching the specific dataset/runtime
+  local label_selector="serverless.fluid.io/inject=true,fluid.io/dataset=${runtime_namespace}.${runtime_name}"
+
+  # Get all pods with the serverless inject label
+  local pods
+  pods=$(kubectl get po -n "${namespace}" -l "${label_selector}" -o jsonpath='{.items[*].metadata.name}' 2>/dev/null)
+  if [[ $? -ne 0 ]]; then
+    echo "failed to get serverless pods in namespace ${namespace}"
+    return
+  fi
+
+  if [[ -n "$pods" ]]; then
+    mkdir -p "$diagnose_dir/pods-${namespace}-serverless"
+    for po in ${pods}; do
+      # Find all containers containing fluid-fuse (covers init-fluid-fuse and fluid-fuse)
+      local containers
+      containers=$(kubectl get po "${po}" -n "${namespace}" -o jsonpath='{.spec.initContainers[*].name} {.spec.containers[*].name}' 2>/dev/null | tr ' ' '\n' | grep 'fluid-fuse')
+      if [[ -n "$containers" ]]; then
+        for container in ${containers}; do
+          kubectl logs "${po}" -c "${container}" -n "${namespace}" &>"$diagnose_dir/pods-${namespace}-serverless/${po}-${container}.log" 2>&1
+        done
+      fi
+    done
+  fi
+  return
+}
+
 core_component() {
   # namespace container selectors...
   local namespace="$1"
@@ -65,7 +109,7 @@ core_component() {
   mkdir -p "$diagnose_dir/pods-${namespace}"
   pods=$(kubectl get po -n ${namespace} "${constrains}" | awk '{print $1}' | grep -v NAME)
   for po in ${pods}; do
-   if [[ "${namespace}"="${fluid_namesapce}" ]]; then
+   if [[ "${namespace}" == "${fluid_namespace}" ]]; then
       kubectl logs "${po}" -c "$container" -n ${namespace} &>"$diagnose_dir/pods-${namespace}/${po}-${container}.log" 2>&1
     else
       kubectl cp "${namespace}/${po}":/opt/goosefs/logs -c "${container}" "$diagnose_dir/pods-${namespace}/${po}-${container}" 2>&1
@@ -93,6 +137,7 @@ pd_collect() {
   pod_status "${fluid_namespace}"
   pod_status "${runtime_namespace}"
   runtime_pod_logs
+  serverless_pod_logs
   fluid_pod_logs
   kubectl_resource
   archive
diff --git a/tools/diagnose-fluid-jindo.sh b/tools/diagnose-fluid-jindo.sh
index 30e3096273c..dd259b13d09 100644
--- a/tools/diagnose-fluid-jindo.sh
+++ b/tools/diagnose-fluid-jindo.sh
@@ -21,7 +21,16 @@ print_usage() {
 run() {
   echo
   echo "-----------------run $*------------------"
-  timeout 10s "$@"
+  if command -v timeout >/dev/null 2>&1; then
+    timeout 10s "$@"
+  elif command -v gtimeout >/dev/null 2>&1; then
+    gtimeout 10s "$@"
+  elif command -v perl >/dev/null 2>&1; then
+    # Use Perl to enforce timeout on systems without GNU coreutils (like standard macOS)
+    perl -e 'alarm shift; exec @ARGV' 10 "$@"
+  else
+    "$@"
+  fi
   if [ $? != 0 ]; then
     echo "failed to collect info: $*"
   fi
@@ -55,6 +64,41 @@ runtime_pod_logs() {
   core_component "${runtime_namespace}" "jindofs-fuse" "role=jindofs-fuse" "release=${runtime_name}"
 }
 
+serverless_pod_logs() {
+  # Check if kubectl is available
+  if ! command -v kubectl >/dev/null 2>&1; then
+    echo "kubectl not found, skipping serverless pod logs collection"
+    return
+  fi
+
+  local namespace="${runtime_namespace}"
+  # More specific selector: injected and matching the specific dataset/runtime
+  local label_selector="serverless.fluid.io/inject=true,fluid.io/dataset=${runtime_namespace}.${runtime_name}"
+
+  # Get all pods with the serverless inject label
+  local pods
+  pods=$(kubectl get po -n "${namespace}" -l "${label_selector}" -o jsonpath='{.items[*].metadata.name}' 2>/dev/null)
+  if [[ $? -ne 0 ]]; then
+    echo "failed to get serverless pods in namespace ${namespace}"
+    return
+  fi
+
+  if [[ -n "$pods" ]]; then
+    mkdir -p "$diagnose_dir/pods-${namespace}-serverless"
+    for po in ${pods}; do
+      # Find all containers containing fluid-fuse (covers init-fluid-fuse and fluid-fuse)
+      local containers
+      containers=$(kubectl get po "${po}" -n "${namespace}" -o jsonpath='{.spec.initContainers[*].name} {.spec.containers[*].name}' 2>/dev/null | tr ' ' '\n' | grep 'fluid-fuse')
+      if [[ -n "$containers" ]]; then
+        for container in ${containers}; do
+          kubectl logs "${po}" -c "${container}" -n "${namespace}" &>"$diagnose_dir/pods-${namespace}-serverless/${po}-${container}.log" 2>&1
+        done
+      fi
+    done
+  fi
+  return
+}
+
 core_component() {
   # namespace container selectors...
   local namespace="$1"
@@ -99,6 +143,7 @@ pd_collect() {
   pod_status "${fluid_namespace}"
   pod_status "${runtime_namespace}"
   runtime_pod_logs
+  serverless_pod_logs
   fluid_pod_logs
   kubectl_resource
   archive
diff --git a/tools/diagnose-fluid-juicefs.sh b/tools/diagnose-fluid-juicefs.sh
index 30d239713ae..e54610c54a6 100644
--- a/tools/diagnose-fluid-juicefs.sh
+++ b/tools/diagnose-fluid-juicefs.sh
@@ -21,7 +21,16 @@ print_usage() {
 run() {
   echo
   echo "-----------------run $*------------------"
-  timeout 10s "$@"
+  if command -v timeout >/dev/null 2>&1; then
+    timeout 10s "$@"
+  elif command -v gtimeout >/dev/null 2>&1; then
+    gtimeout 10s "$@"
+  elif command -v perl >/dev/null 2>&1; then
+    # Use Perl to enforce timeout on systems without GNU coreutils (like standard macOS)
+    perl -e 'alarm shift; exec @ARGV' 10 "$@"
+  else
+    "$@"
+  fi
   if [ $? != 0 ]; then
     echo "failed to collect info: $*"
   fi
@@ -54,6 +63,41 @@ runtime_pod_logs() {
   core_component "${runtime_namespace}" "juicefs-fuse" "role=juicefs-fuse" "release=${runtime_name}"
 }
 
+serverless_pod_logs() {
+  # Check if kubectl is available
+  if ! command -v kubectl >/dev/null 2>&1; then
+    echo "kubectl not found, skipping serverless pod logs collection"
+    return
+  fi
+
+  local namespace="${runtime_namespace}"
+  # More specific selector: injected and matching the specific dataset/runtime
+  local label_selector="serverless.fluid.io/inject=true,fluid.io/dataset=${runtime_namespace}.${runtime_name}"
+
+  # Get all pods with the serverless inject label
+  local pods
+  pods=$(kubectl get po -n "${namespace}" -l "${label_selector}" -o jsonpath='{.items[*].metadata.name}' 2>/dev/null)
+  if [[ $? -ne 0 ]]; then
+    echo "failed to get serverless pods in namespace ${namespace}"
+    return
+  fi
+
+  if [[ -n "$pods" ]]; then
+    mkdir -p "$diagnose_dir/pods-${namespace}-serverless"
+    for po in ${pods}; do
+      # Find all containers containing fluid-fuse (covers init-fluid-fuse and fluid-fuse)
+      local containers
+      containers=$(kubectl get po "${po}" -n "${namespace}" -o jsonpath='{.spec.initContainers[*].name} {.spec.containers[*].name}' 2>/dev/null | tr ' ' '\n' | grep 'fluid-fuse')
+      if [[ -n "$containers" ]]; then
+        for container in ${containers}; do
+          kubectl logs "${po}" -c "${container}" -n "${namespace}" &>"$diagnose_dir/pods-${namespace}-serverless/${po}-${container}.log" 2>&1
+        done
+      fi
+    done
+  fi
+  return
+}
+
 core_component() {
   # namespace container selectors...
   local namespace="$1"
@@ -98,6 +142,7 @@ pd_collect() {
   pod_status "${fluid_namespace}"
   pod_status "${runtime_namespace}"
   runtime_pod_logs
+  serverless_pod_logs
   fluid_pod_logs
   kubectl_resource
   archive