如何将时间序列CSV转换为Prometheus的OpenMetrics格式？

针对你提供的CSV结构（包含时间、站点ID和4个传感器值），可以通过以下两种可靠方式完成格式转换：

方法一：自定义Python脚本（最灵活可控） #

如果工具无法适配你的数据格式，自己写脚本是最直接的解决方案。以下脚本会将CSV逐行转换为符合OpenMetrics规范的内容：

import csv
from datetime import datetime

# 定义每个传感器对应的Prometheus指标元数据
metric_defs = {
    "s0": {"name": "station_sensor_s0", "help": "Reading from sensor s0 at weather station", "type": "gauge"},
    "s1": {"name": "station_sensor_s1", "help": "Reading from sensor s1 at weather station", "type": "gauge"},
    "s2": {"name": "station_sensor_s2", "help": "Reading from sensor s2 at weather station", "type": "gauge"},
    "s3": {"name": "station_sensor_s3", "help": "Reading from sensor s3 at weather station", "type": "gauge"},
}

# 初始化OpenMetrics头部（必须包含HELP和TYPE注释）
output_lines = []
for metric in metric_defs.values():
    output_lines.append(f"# HELP {metric['name']} {metric['help']}")
    output_lines.append(f"# TYPE {metric['name']} {metric['type']}")

# 读取CSV并转换每一行数据
with open("your_data.csv", "r", encoding="utf-8") as csv_file:
    reader = csv.DictReader(csv_file)
    for row in reader:
        # 将ISO格式时间转换为Unix时间戳（秒级，Prometheus默认支持）
        unix_ts = int(datetime.fromisoformat(row["time"]).timestamp())
        station_id = row["id_station"]
        
        # 为每个传感器生成指标行
        for sensor_col in ["s0", "s1", "s2", "s3"]:
            metric_info = metric_defs[sensor_col]
            sensor_value = row[sensor_col]
            output_lines.append(f'{metric_info["name"]}{{id_station="{station_id}"}} {sensor_value} {unix_ts}')

# 将结果写入输出文件
with open("station_metrics.openmetrics", "w", encoding="utf-8") as out_file:
    out_file.write("\n".join(output_lines) + "\n")

使用说明： #

1. 将脚本中的your_data.csv替换为你的CSV文件路径
2. 运行脚本后会生成station_metrics.openmetrics文件
3. 可以用Prometheus的promtool验证格式正确性：

   promtool check metrics station_metrics.openmetrics
   

方法二：调整csv-exporter配置（适配现有工具） #

如果坚持使用csv-exporter，问题大概率出在配置文件未匹配你的CSV结构。以下是针对你的数据的示例配置：

jobs:
  - name: "station_sensors_export"
    url: "file:///path/to/your_data.csv"
    metrics:
      - name: "station_sensor_s0"
        type: "gauge"
        help: "Sensor s0 reading"
        labels:
          id_station: "{{.id_station}}"
        value: "{{.s0}}"
        timestamp: "{{.time}}"
        timestamp_format: "2006-01-02T15:04:05"
      - name: "station_sensor_s1"
        type: "gauge"
        help: "Sensor s1 reading"
        labels:
          id_station: "{{.id_station}}"
        value: "{{.s1}}"
        timestamp: "{{.time}}"
        timestamp_format: "2006-01-02T15:04:05"
      - name: "station_sensor_s2"
        type: "gauge"
        help: "Sensor s2 reading"
        labels:
          id_station: "{{.id_station}}"
        value: "{{.s2}}"
        timestamp: "{{.time}}"
        timestamp_format: "2006-01-02T15:04:05"
      - name: "station_sensor_s3"
        type: "gauge"
        help: "Sensor s3 reading"
        labels:
          id_station: "{{.id_station}}"
        value: "{{.s3}}"
        timestamp: "{{.time}}"
        timestamp_format: "2006-01-02T15:04:05"

使用说明： #

1. 将配置中的文件路径替换为你的CSV路径
2. 启动csv-exporter时指定该配置文件
3. 通过http://localhost:9275/metrics访问导出的指标，保存为文件或直接配置Prometheus scrape该端点

内容的提问来源于stack exchange，提问作者AbdelKh