[fix][mds] Make crontab lifecycle safe; simplify scheduling

.agents/skills/dev-deploy/SKILL.md

@@ -18,7 +18,7 @@ context: fork
 - `SERVER_HOST` / `SERVER_LISTEN_HOST`：对外 / 监听地址
 - `SERVER_START_PORT`：起始端口；第 `i` 个实例为 `SERVER_START_PORT + i`，即首实例 `7801`
 - `CLUSTER_ID`、`MDS_INSTANCE_START_ID`、`COORDINATOR_ADDR`
-- `STORAGE_ENGINE` / `STORAGE_URL`：由 `deploy_mds.sh` 写进 `mds.conf`
+- `STORAGE_ENGINE` / `STORAGE_URL`：由 `operate_mds.sh` 写进 `mds.conf`
 - `S3_ENDPOINT` / `S3_AK` / `S3_SK` / `S3_BUCKETNAME`、`LOCAL_DATASTORE_PATH`：仅 `create_fs.sh` 使用
 
 ## 步骤
@@ -45,7 +45,7 @@ context: fork
 3. **部署并启动 MDS**
 
    ```bash
-   bash clean_start.sh --server_num=$SERVER_NUM
+   bash operate_mds.sh restart --server_num=$SERVER_NUM
    ```
 
    判据：`pgrep -c dingo-mds` 输出等于 `SERVER_NUM`。（不要用 `ps -ef | grep`，它会匹配到自己，也不校验数量。）
@@ -95,4 +95,13 @@ bash create_fs.sh --fs_name=$FS_NAME --mds_addr=$SERVER_HOST:$(($SERVER_START_PO
 
 ## 其他脚本
 
-`deploy_mds.sh` / `start_mds.sh` / `stop_mds.sh` 是 `clean_start.sh` 的拆分，只重启 MDS 时单独用。参数以 `bash <脚本> --help` 为准。
+`operate_mds.sh` 一个脚本管全部 MDS 生命周期，子命令选动作，flag 两种写法都行（`--server_num=2 restart` 同 `restart --server_num=2`）：
+
+```bash
+bash operate_mds.sh restart --server_num=1   # = stop + deploy + start，等价于原 clean_start.sh
+bash operate_mds.sh stop --server_num=1      # 停；--force 用 kill -9，--use_pgrep 按进程名而非 pid 文件
+bash operate_mds.sh deploy --server_num=1    # 只重新部署（软链二进制 + 渲染 conf）
+bash operate_mds.sh start --server_num=1     # 只启动
+```
+
+参数以 `bash operate_mds.sh --help` 为准。`restart` 中 stop / deploy 失败会直接中止，不会带着半坏状态继续起服务。

.github/scripts/deploy-mds-client.sh

@@ -18,6 +18,8 @@ CLUSTER_ID=101
 MDS_PORT=8821
 MDS_INSTANCE_ID=1001
 VFS_DUMMY_PORT=10001
+MDS_LOG_LEVEL=DEBUG
+MDS_LOG_V=20
 
 BUILD_DIR="${GITHUB_WORKSPACE}/build/bin"
 TEMPLATE="${GITHUB_WORKSPACE}/scripts/dev-mds/mds.template.conf"
@@ -54,6 +56,8 @@ sed -i "s|\\\$SERVER_PORT|${MDS_PORT}|g"           "${DIST_CONF}"
 sed -i "s|\\\$BASE_PATH|${MDS_DIST}|g"             "${DIST_CONF}"
 sed -i "s|\\\$STORAGE_ENGINE|dingo-store|g"        "${DIST_CONF}"
 sed -i "s|\\\$STORAGE_URL|list://${COORDINATOR_ADDR}|g" "${DIST_CONF}"
+sed -i "s|\\\$LOG_LEVEL|${MDS_LOG_LEVEL}|g"          "${DIST_CONF}"
+sed -i "s|\\\$LOG_V|${MDS_LOG_V}|g"                  "${DIST_CONF}"
 
 echo "${COORDINATOR_ADDR}" > "${MDS_DIST}/conf/coor_list"
 

.gitignore

@@ -115,3 +115,6 @@ bench_results/
 
 CONTEXT.md
 .pi-subagents/
+
+# Local dev config (contains secrets)
+scripts/dev-mds/env.local

scripts/deploy/deploy.sh

@@ -76,6 +76,8 @@ function deploy_server() {
     sed  -i 's,\$BASE_PATH,'"$dstpath"',g'                        $dist_conf
     sed  -i 's,\$STORAGE_ENGINE,dingo-store,g'                    $dist_conf
     sed  -i 's,\$STORAGE_URL,file://./conf/coor_list,g'           $dist_conf
+    sed  -i 's,\$LOG_LEVEL,INFO,g'                                $dist_conf
+    sed  -i 's,\$LOG_V,0,g'                                       $dist_conf
 
     # coor_list file
     coor_file="${dstpath}/conf/coor_list"

scripts/dev-mds/create_cluster.sh

@@ -5,7 +5,7 @@ if [[ ! -d "$mydir" ]]; then mydir="$PWD"; fi
 . $mydir/shflags
 
 DEFINE_integer cluster_id 0 'cluster id'
-DEFINE_string parameters 'mds_deploy_parameters.local' 'deploy parameters file'
+DEFINE_string env 'env.local' 'deploy env file'
 
 
 # parse the command-line
@@ -14,7 +14,7 @@ eval set -- "${FLAGS_ARGV}"
 
 echo "cluster_id: ${FLAGS_cluster_id}"
 
-source $mydir/${FLAGS_parameters}
+source $mydir/${FLAGS_env}
 
 
 #check cluster id is valid

scripts/dev-mds/create_fs.sh

@@ -6,13 +6,19 @@ if [[ ! -d "$mydir" ]]; then mydir="$PWD"; fi
 
 DEFINE_string fs_name '' 'fs name'
 DEFINE_string mds_addr '' 'mds address'
-DEFINE_string parameters 'mds_deploy_parameters.local' 'deploy parameters file'
+DEFINE_string env 'env.local' 'deploy env file'
 DEFINE_boolean use_local_datastore false 'use local datastore'
 
 # parse the command-line
 FLAGS "$@" || exit 1
 eval set -- "${FLAGS_ARGV}"
 
+source $mydir/${FLAGS_env}
+
+if [ -z "${FLAGS_mds_addr}" ]; then
+    FLAGS_mds_addr="${SERVER_HOST}:$((SERVER_START_PORT + 1))"
+fi
+
 echo "fs_name: ${FLAGS_fs_name}"
 echo "mds_addr: ${FLAGS_mds_addr}"
 
@@ -27,8 +33,6 @@ if [ -z "${FLAGS_mds_addr}" ]; then
     exit -1
 fi
 
-source $mydir/${FLAGS_parameters}
-
 BASE_DIR=$(dirname $(dirname $(cd $(dirname $0); pwd)))
 BUILD_DIR=$BASE_DIR/build
 MDS_CLIENT_BIN_PATH=$BUILD_DIR/bin/dingo-mds-client

scripts/dev-mds/mds_deploy_parameters → scripts/dev-mds/env.example

@@ -12,4 +12,10 @@ STORAGE_URL=list://127.0.0.1:2379
 S3_ENDPOINT=
 S3_AK=
 S3_SK=
-S3_BUCKETNAME=
+S3_BUCKETNAME=
+
+
+LOCAL_DATASTORE_PATH=
+
+LOG_LEVEL=DEBUG
+LOG_V=0

scripts/dev-mds/mds.template.conf

@@ -27,8 +27,8 @@
 
 # log
 --log_dir=$BASE_PATH/log
---log_level=DEBUG
---log_v=0
+--log_level=$LOG_LEVEL
+--log_v=$LOG_V
 
 
 # dingo-sdk