Clarification on `get_metric_range_data` method

[kev-pebble]

I am querying the Prometheus Host. I specify the start, end time and chunksize as 1 hour. When querying the data between the start and end times with 1 hour chunksize, is the data returned aggregated sum by the hour or simply the value at the hourly timestamp? My metric is a counter metric like kepler_container_joules_total, so the returned values as cumulative totals at the specified hourly timestamps?

def query_data(victoria_db_url, start_time, end_time, metric_name=None):

    prom = PrometheusConnect(url=victoria_db_url, disable_ssl=True)
    
    # Run query
    result = prom.all_metrics()
    
    kepler_metrics = [metric for metric in result if 'kepler' in metric]
    
    # save dfs as list
    df_list = []

    # set default metric
    if metric_name == None:
        metric_name == "kepler_container_joules_total"
        
    # Query the data
    for metric in kepler_metrics:

        if metric == metric_name:
        
            #print(f"Metric: {metric}")
            data = prom.get_metric_range_data(
                                                metric_name=metric,
                                                start_time = start_time,
                                                end_time = end_time,
                                                chunk_size = timedelta(hours=1)
                                             )
            
            if data:
                # Convert the metric data to a DataFrame
                metric_df = MetricSnapshotDataFrame(data)
                # Append the DataFrame to the list
                df_list.append(metric_df)

    df_kepler = pd.concat(df_list, ignore_index=True)

    return df_kepler

Is there a way to get aggregate hourly power consumption without the cumulative / counter time series data?

@4n4nd

[4n4nd]

When querying the data between the start and end times with 1 hour chunksize, is the data returned aggregated sum by the hour or simply the value at the hourly timestamp?

chunk_size is just for query chunks. For example, if you request 6 hours of data (let's say 12pm to 6pm) and keep chunk_size to 1 hour, the client will request data from prometheus in 1 hour chunks. 12pm-1pm, 1pm-2pm, 2pm-.....

There is no aggregation performed by the function get_metric_range_data.

Is there a way to get aggregate hourly power consumption without the cumulative / counter time series data?

you might need to check promQL docs to see if you can write up a query that does it for you.

[kev-pebble]

Ok - got it.

So, the value returned by get_metric_range_data function for 6 hours of requested data between 12-1pm will be the observations at that hourly timestamp without any aggregations?

[4n4nd]

Well not really. It will still get you the raw data from prometheus for your query. No aggregations.

So, the data collected using chunk_size = 6 hours from time range 12pm to 6pm should be the same as data collected using chunk_size = 1hour from time range 12pm to 6pm.

The chunk_size parameter is to break your query response into smaller chunks because sometimes you can have too much data come in and that can cause problems