Pandas解决不同指标变量采集周期不一的问题（resample+merge时间戳拼接）

背景

我有好几台服务器， power（时间戳为unix） , cpu在不同的系统中采集， 而且他们的采集周期不一；现需要把sn, power, cpu拼接在同一个 df中；
具体如下：Pandas resample把时间戳变为相同的采集间隔，然后根据时间戳进行merge拼接

cpu数据如下(cpu_utils.csv)：

,agent_sn,cpu_percent_in,time
2022-02-24 08:40:57.113353+00:00,123456,99,2022-02-24 08:40:57.113353
2022-02-22 23:33:00.000000+00:00,123456,99,2022-09-22 23:33:00.000000
2022-09-22 23:33:00+00:00,123456,99,2022-09-22 15:33:00.000000
2022-09-24 06:47:49.409795+00:00,123456,99,2022-09-24 06:47:49.409795
2022-09-24 06:48:21.449992+00:00,123456,99,2022-09-24 06:48:21.449992
2022-09-24 06:48:55.841323+00:00,123456,99,2022-09-24 06:48:55.841323
2022-09-24 06:49:01.838624+00:00,123456,99,2022-09-24 06:49:01.838624
2022-09-24 07:48:29.881619+00:00,123456,88,2022-09-24 07:48:29.881619
2022-09-24 07:48:39.119652+00:00,123456,88,2022-09-24 07:48:39.119652

resample+merge代码

if __name__ == "__main__":

    dict_power = {"1663889620000": "390", "1663889640000": "348", "1663889660000": "378",
                  "1663889680000": "378", "1663889700000": "372", "1663889740000": "372",
                  "1663889760000": "366", "1663889810000": "384", "1663889840000": "378",
                  "1663889860000": "390", "1663889870000": "384", "1663889880000": "378",
                  "1663889900000": "384", "1663889940000": "360", "1663889970000": "360",
                  "1663889980000": "402", "1663890020000": "384", "1663890030000": "372",
                  "1663890050000": "384", "1663890070000": "378", "1663890080000": "360",
                  "1663890120000": "360", "1663890140000": "360", "1663890160000": "366",
                  "1663890180000": "372", "1663890210000": "378", "1663890220000": "426",
                  "1663890260000": "354", "1663890280000": "372", "1663890290000": "360",
                  "1663890300000": "378", "1663890310000": "444", "1663890320000": "384",
                  "1663890330000": "366", "1663890360000": "360", "1663890380000": "378",
                  "1663890390000": "384", "1663890400000": "378", "1663890420000": "354",
                  "1663890430000": "378", "1663890450000": "372", "1663890490000": "384",
                  "1663890530000": "378", "1663890550000": "390", "1663890600000": "390",
                  "1663890610000": "366", "1663890630000": "348", "1663890640000": "360",
                  "1663890660000": "384", "1663890670000": "348", "1663890690000": "360",
                  "1663890740000": "384", "1663890760000": "0", "1663890800000": "378"}

    sn = "123456"
    result = pd.DataFrame()
    new_df = pd.DataFrame()
    # dict_power = value
    tss = [int(x) for x in list(dict_power.keys())]
    powers = [int(x) for x in list(dict_power.values())]
    test_df = pd.DataFrame(data={"ts": tss, "power": powers, "sn": [sn] * len(dict_power)})
    merge_data = test_df
    df_temp = pd.DataFrame()
    df_temp["power"] = pd.Series(dtype="object")
    df_temp.at[0, "power"] = merge_data["power"].values
    df_temp["sn"] = merge_data["sn"].iloc[0]
    # print("df_temp", df_temp)
    test_df['ts'] = pd.to_datetime(test_df['ts'], unit='ms')
    test_df = test_df.set_index("ts")
    new_df["power"] = test_df.power.resample(rule='1T').mean()
    new_df["sn"] = sn

    result = pd.concat([result, new_df.reset_index()])
    print("result", result)
    cpu_utils = pd.read_csv("../cpu_utils.csv")
    cpu_utils["time"] = pd.to_datetime(cpu_utils["time"])
    # print("cpu_utils", pd.to_datetime(cpu_utils["time"], format="%Y-%m-%d",exact="strict"))
    print("cpu_utils", cpu_utils)

    merge_data = pd.merge(cpu_utils, result, how='inner', left_on="time", right_on="ts")
    print("merge_data", merge_data)