Very long waiting times for S2S retrieval requests on ECDS

Datasets and Usage
1

Hello everyone,

I have recently experienced extremely long waiting times when downloading S2S forecast data from ECDS using the CDS API.

For example:

  • A request submitted on May 22 took approximately 21 hours before it finally completed successfully.

  • The same request submitted again on May 25 has still not completed after waiting for many hours.

Below is the Python code I am using:

import cdsapi

STEP_6H = [str(6*i) for i in range(1, 185)]

dataset = "s2s-forecasts"

request = {
    "origin": "ecmwf",
    "year": "2026",
    "month": "05",
    "day": "19",
    "time": "00:00",
    "level_type": "single_level",
    "variable": [
        "10_m_u_component_of_wind",
        "10_m_v_component_of_wind",
        "maximum_2_m_temperature_in_the_last_6_hours",
        "mean_sea_level_pressure",
        "minimum_2_m_temperature_in_the_last_6_hours",
        "surface_pressure",
        "total_precipitation"
    ],
    "forecast_type": "control_forecast",
    "leadtime_hour": STEP_6H,
    "data_format": "grib",
    "area": [25.7, 109.4, 20, 118],
}

target = "test.grib"

client = cdsapi.Client(key=key, url=url)
client.retrieve(dataset, request, target)

I would also like to know whether this issue is caused by my request configuration or by the ECDS server side.

The requested data volume is actually not very large — the final GRIB file is only a few tens of MB.

So I am wondering whether:

  • there is something inefficient in my request setup,

  • or whether the long waiting time is mainly due to current server load and queue congestion.

Any advice would be appreciated.

2

Hello.

I too have noticed very long processing time compared to past queries ( though for reanalyses, not forecast data ).

Have you learned the reasons for this?

3

I encountered the same problem as you. The download speed is very slow, and I have to queue for a long time. Has anyone solved this problem?

4

No even for reanalysis it is the same problem it is just running for hours and hours even for a simple monthly data.