feat: add nginx reload and validation wrappers with sudo permissions

apps/agent/docker/s6/cont-init.d/20-install-reload-wrapper

@@ -0,0 +1,170 @@
+#!/bin/sh
+set -eu
+
+# This init script installs a minimal nginx reload wrapper and a sudoers
+# entry so the `yanpm-agent` user can perform a controlled reload via sudo.
+
+WRAPPER_PATH="/usr/local/sbin/yanpm-nginx-reload"
+SUDOERS_PATH="/etc/sudoers.d/yanpm-agent"
+AGENT_USER="${YANPM_AGENT_USER:-yanpm-agent}"
+
+# validate wrapper
+VALIDATE_PATH="/usr/local/sbin/yanpm-nginx-validate"
+# validate file wrapper
+VALIDATE_FILE_PATH="/usr/local/sbin/yanpm-nginx-validate-file"
+
+echo "[cont-init.d] install-reload-wrapper: setting up nginx reload helper"
+
+# find nginx binary
+NGINX_BIN="$(command -v nginx || true)"
+if [ -z "${NGINX_BIN}" ]; then
+  echo "Warning: nginx binary not found in PATH; wrapper will still be created but may fail at runtime." >&2
+  NGINX_BIN="/usr/sbin/nginx"
+fi
+
+# Create wrapper
+mkdir -p /usr/local/sbin /etc/sudoers.d
+
+cat > "${WRAPPER_PATH}" <<- 'EOF'
+#!/bin/sh
+exec "@NGINX_BIN@" -c /etc/nginx/nginx.conf -s reload
+EOF
+
+# Replace placeholder with actual path
+sed -i "s|@NGINX_BIN@|${NGINX_BIN}|g" "${WRAPPER_PATH}" || true
+
+chmod 0750 "${WRAPPER_PATH}"
+chown root:root "${WRAPPER_PATH}" || true
+
+#
+#
+#
+
+# Create validate wrapper
+cat > "${VALIDATE_PATH}" <<- 'EOF'
+#!/bin/sh
+exec "@NGINX_BIN@" -c /etc/nginx/nginx.conf -t
+EOF
+
+# Replace placeholder with actual path in validate wrapper
+sed -i "s|@NGINX_BIN@|${NGINX_BIN}|g" "${VALIDATE_PATH}" || true
+
+chmod 0750 "${VALIDATE_PATH}"
+chown root:root "${VALIDATE_PATH}" || true
+
+#
+#
+#
+
+# Create validate file wrapper (secure)
+cat > "${VALIDATE_FILE_PATH}" <<-'EOF'
+#!/bin/sh
+set -eu
+
+if [ $# -ne 1 ]; then
+  echo "Usage: $0 <nginx-config-file>" >&2
+  exit 2
+fi
+
+INPUT="$1"
+
+# Resolve absolute path
+if command -v readlink >/dev/null 2>&1; then
+  TARGET="$(readlink -f -- "$INPUT" 2>/dev/null || true)"
+elif command -v realpath >/dev/null 2>&1; then
+  TARGET="$(realpath -- "$INPUT" 2>/dev/null || true)"
+else
+  echo "Error: no path resolver (readlink/realpath) available" >&2
+  exit 3
+fi
+
+if [ -z "$TARGET" ]; then
+  echo "Error: cannot resolve path: $INPUT" >&2
+  exit 4
+fi
+
+# Must be a regular file and not a symlink
+if [ ! -f "$TARGET" ] || [ -L "$TARGET" ]; then
+  echo "Error: ${TARGET} is not a regular file" >&2
+  exit 5
+fi
+
+# must be created by agent user
+AGENT_UID="$(id -u yanpm-agent 2>/dev/null || true)"
+if [ -z "$AGENT_UID" ]; then
+  echo "Error: yanpm-agent user not found" >&2
+  exit 6
+fi
+
+FILE_UID="$(stat -c %u -- "$TARGET" 2>/dev/null || true)"
+if [ "$FILE_UID" != "$AGENT_UID" ]; then
+  echo "Error: ${TARGET} not owned by yanpm-agent user" >&2
+  exit 7
+fi
+
+# Ensure file is not world-writable; allow typical 664 (rw-rw-r--)
+if command -v stat >/dev/null 2>&1; then
+  MODE="$(stat -c %a -- "$TARGET" 2>/dev/null || true)"
+  if [ -n "$MODE" ]; then
+    OTHERS=$(( MODE % 10 ))
+    if [ $(( OTHERS & 2 )) -ne 0 ]; then
+      echo "Error: ${TARGET} is world-writable" >&2
+      exit 8
+    fi
+  fi
+elif command -v find >/dev/null 2>&1; then
+  if find "$TARGET" -maxdepth 0 -perm /002 -print -quit >/dev/null 2>&1; then
+    echo "Error: ${TARGET} is world-writable" >&2
+    exit 8
+  fi
+fi
+
+exec "@NGINX_BIN@" -c "$TARGET" -t
+EOF
+
+# Replace placeholder with actual path in validate file wrapper
+sed -i "s|@NGINX_BIN@|${NGINX_BIN}|g" "${VALIDATE_FILE_PATH}" || true
+chmod 0750 "${VALIDATE_FILE_PATH}"
+chown root:root "${VALIDATE_FILE_PATH}" || true
+
+echo "Created wrapper: ${WRAPPER_PATH} (owned by root, mode 750)"
+
+#
+#
+#
+
+# Ensure sudoers entry exists allowing the agent to run only this wrapper as root
+if command -v sudo >/dev/null 2>&1; then
+  echo "sudo present; creating sudoers entry"
+  cat > "${SUDOERS_PATH}" <<- EOF
+# Allow ${AGENT_USER} to run the nginx reload and validate wrappers without a password
+${AGENT_USER} ALL=(root) NOPASSWD: ${WRAPPER_PATH}, ${VALIDATE_PATH}, ${VALIDATE_FILE_PATH}
+EOF
+  chmod 0440 "${SUDOERS_PATH}" || true
+  echo "Wrote sudoers entry: ${SUDOERS_PATH}"
+else
+  echo "sudo not found; attempting to install"
+  if command -v apk >/dev/null 2>&1; then
+    apk add --no-cache sudo || true
+  elif command -v apt-get >/dev/null 2>&1; then
+    apt-get update || true
+    apt-get install -y sudo || true
+  elif command -v yum >/dev/null 2>&1; then
+    yum install -y sudo || true
+  else
+    echo "No known package manager to install sudo; please ensure sudo is available in the image." >&2
+  fi
+
+  if command -v sudo >/dev/null 2>&1; then
+    cat > "${SUDOERS_PATH}" <<- EOF
+# Allow ${AGENT_USER} to run the nginx reload and validate wrappers without a password
+${AGENT_USER} ALL=(root) NOPASSWD: ${WRAPPER_PATH}, ${VALIDATE_PATH}, ${VALIDATE_FILE_PATH}
+EOF
+    chmod 0440 "${SUDOERS_PATH}" || true
+    echo "Installed sudo and wrote sudoers entry: ${SUDOERS_PATH}"
+  else
+    echo "Failed to install sudo; the agent will not be able to reload nginx via sudo." >&2
+  fi
+fi
+
+exit 0

apps/agent/justfile

@@ -0,0 +1,2 @@
+build-docker:
+  docker build -t yanpm/agent:latest .

apps/agent/src/commands/reload.rs

@@ -93,6 +93,17 @@ impl ReloadCommand {
         }
 
         // reload the running nginx master process (no -c) so it reloads its configured main config
+        // Prefer the restricted sudo wrapper if available, fall back to direct nginx reload.
+        // TODO: allow configuring the path to the wrapper
+        match run_cmd("sudo", &["-n", "/usr/local/sbin/yanpm-nginx-reload"], 10).await {
+            Ok(res) => Ok(res),
+            Err(e) => {
+                error!(
+                    "sudo reload wrapper failed, falling back to direct nginx reload: {}",
+                    e
+                );
                 run_cmd("nginx", &["-s", "reload"], 10).await
             }
         }
+    }
+}

apps/agent/src/commands/validate.rs

@@ -1,3 +1,5 @@
+use tracing::{error, info, warn};
+
 use crate::commands::{run::run_cmd, write_config::INTERNAL_CONFIG_FOLDER_NAME};
 use std::path::PathBuf;
 
@@ -17,7 +19,51 @@ impl ValidateCommand {
     pub async fn validate_all(
         &self,
     ) -> Result<(i32, String), Box<dyn std::error::Error + Send + Sync>> {
-        run_cmd("nginx", &["-t"], 10).await
+        // Try a normal config test first. If it fails due to pid permission
+        // errors (common when running unprivileged against /run/nginx.pid),
+        // retry with a writable pid override so validation can succeed.
+        match run_cmd("nginx", &["-t"], 10).await {
+            Ok(res) => Ok(res),
+            Err(e) => {
+                info!(
+                    "nginx -t failed: {}. Trying with privileged wrapper or writable pid override.",
+                    e
+                );
+                let es = e.to_string();
+                if es.contains("/run/nginx.pid") && es.contains("Permission denied") {
+                    // Try privileged validate wrapper if available (allows the agent to run
+                    // nginx -t via sudo without modifying the main config).
+                    match run_cmd(
+                        "sudo",
+                        // TODO: allow configuring the path to the wrapper
+                        &["-n", "/usr/local/sbin/yanpm-nginx-validate"],
+                        10,
+                    )
+                    .await
+                    {
+                        Ok(res) => return Ok(res),
+                        Err(e) => {
+                            warn!(
+                                "Privileged validate wrapper failed: {}. Falling back to writable pid override.",
+                                e
+                            );
+                            // Fallback to the existing writable-pid override if sudo wrapper
+                            // isn't available or fails.
+                            let pid_path = format!(
+                                "{}/yanpm-validate-{}.pid",
+                                std::env::temp_dir().display(),
+                                std::process::id()
+                            );
+                            let g_arg = format!("pid {};", pid_path);
+                            let args_vec = ["-t".to_string(), "-g".to_string(), g_arg];
+                            let args_ref: Vec<&str> = args_vec.iter().map(|s| s.as_str()).collect();
+                            return run_cmd("nginx", args_ref.as_slice(), 10).await;
+                        }
+                    }
+                }
+                Err(e)
+            }
+        }
     }
 
     pub async fn validate(
@@ -37,11 +83,84 @@ impl ValidateCommand {
             return Err(format!("Config file not found: {}", full_path.display()).into());
         }
 
-        run_cmd(
+        // Create a temporary wrapper nginx config that provides the required
-            "nginx",
+        // top-level sections (`events` and `http`) and includes the fragment.
-            &["-t", "-c", full_path.to_str().ok_or("invalid config path")?],
+        let fragment_path = full_path.to_str().ok_or("invalid config path")?.to_string();
+
+        let mut tmp_path = std::env::temp_dir();
+        let tmp_name = format!("yanpm-validate-{}-{}.conf", timestamp, std::process::id());
+        tmp_path.push(tmp_name);
+
+        let wrapper = format!(
+            "worker_processes 1;\nevents {{ worker_connections 1024; }}\nhttp {{\n    include {};\n}}\n",
+            fragment_path
+        );
+
+        // Write the temporary wrapper file
+        tokio::fs::write(&tmp_path, wrapper).await?;
+        let tmp_path_str = tmp_path
+            .to_str()
+            .ok_or("invalid temp config path")?
+            .to_string();
+
+        // Run the test against the wrapper, telling nginx to place its pid
+        // somewhere writable so the config test doesn't fail with permission
+        // errors when running as an unprivileged user.
+        let result = match run_cmd("nginx", &["-t", "-c", &tmp_path_str], 10).await {
+            Ok(res) => Ok(res),
+            Err(e) => {
+                info!(
+                    "nginx -t failed: {}. Trying with privileged wrapper or writable pid override.",
+                    e
+                );
+                let es = e.to_string();
+                if es.contains("/run/nginx.pid") && es.contains("Permission denied") {
+                    // Try privileged validate wrapper if available (allows the agent to run
+                    // nginx -t via sudo without modifying the main config).
+                    match run_cmd(
+                        "sudo",
+                        // TODO: allow configuring the path to the wrapper
+                        &[
+                            "-n",
+                            "/usr/local/sbin/yanpm-nginx-validate-file",
+                            &tmp_path_str,
+                        ],
                         10,
                     )
                     .await
+                    {
+                        Ok(res) => return Ok(res),
+                        Err(e) => {
+                            warn!(
+                                "Privileged validate wrapper failed: {}. Falling back to writable pid override.",
+                                e
+                            );
+                            let pid_path = format!(
+                                "{}/yanpm-validate-{}.pid",
+                                std::env::temp_dir().display(),
+                                std::process::id()
+                            );
+                            let g_arg = format!("pid {};", pid_path);
+
+                            let args_vec = [
+                                "-t".to_string(),
+                                "-c".to_string(),
+                                tmp_path_str.clone(),
+                                "-g".to_string(),
+                                g_arg,
+                            ];
+                            let args_ref: Vec<&str> = args_vec.iter().map(|s| s.as_str()).collect();
+
+                            return run_cmd("nginx", args_ref.as_slice(), 10).await;
+                        }
+                    }
+                }
+                Err(e)
+            }
+        };
+
+        let _ = tokio::fs::remove_file(&tmp_path).await;
+
+        result
     }
 }

apps/agent/src/commands/write_config.rs

@@ -2,6 +2,7 @@ use std::os::unix::fs::PermissionsExt;
 use std::path::PathBuf;
 use std::time::{SystemTime, UNIX_EPOCH};
 use tokio::io::AsyncWriteExt;
+use tracing::info;
 
 use crate::commands::run::to_file_name;
 
@@ -26,6 +27,7 @@ impl WriteConfigCommand {
         let path = self.nginx_config_dir.clone();
         // ensure main config dir exists
         tokio::fs::create_dir_all(&path).await?;
+        info!("Writing config to {:?}", path.join(&filename));
 
         // create YANPM subdir where fragment files live
         let yanpm_dir = path.join(INTERNAL_CONFIG_FOLDER_NAME);
@@ -62,7 +64,7 @@ impl WriteConfigCommand {
 
         // set explicit permissions (rw-r-----)
         tokio::fs::set_permissions(&final_path, std::fs::Permissions::from_mode(0o640)).await?;
-
+        info!("Config written and permissions set for {:?}", final_path);
         Ok(())
     }
 }

apps/agent/src/routes.rs

@@ -43,7 +43,7 @@ pub async fn validate(
         }
     };
 
-    let (_code, _output) = match nginx_controller
+    let resp = match nginx_controller
         .validate(&params.config_name, params.timestamp)
         .await
     {
@@ -54,7 +54,7 @@ pub async fn validate(
         }
     };
 
-    (axum::http::StatusCode::OK,).into_response()
+    (axum::http::StatusCode::OK, axum::Json(resp)).into_response()
 }
 
 #[derive(Deserialize)]

apps/container/src/main.rs

@@ -20,6 +20,12 @@ struct Args {
     db_type: String,
 
     // agent related
+    /// agent image name
+    #[arg(long, default_value = "yanpm/agent", env = "AGENT_IMAGE_NAME")]
+    agent_image: String,
+    /// agent image tag
+    #[arg(long, default_value = "latest", env = "AGENT_IMAGE_TAG")]
+    agent_image_tag: String,
     /// force build agent image
     #[arg(long, default_value_t = false, env = "AGENT_FORCE_BUILD")]
     agent_force_build: bool,
@@ -172,8 +178,8 @@ async fn parse_args() -> ParsedArgs {
         db_type: args.db_type,
         agent_container_config: Some(AgentContainerConfig {
             // TODO: allow customization of these fields via CLI args
-            image: "yanpm-agent".to_string(),
+            image: args.agent_image,
-            tag: "latest".to_string(),
+            tag: args.agent_image_tag,
             container_name: format!("yanpm-agent-container-{}", time),
             dockerfile_path,
             force_build: args.agent_force_build,