ERA5-CDSAPI-Extremely Slow download

Ever since the update downloading data from ERA5 has become extremely slow and waiting time increased from minutes to days. Any chance this change for the better? I need 60 years of data (june/july/august) for following variables (T2M/PRCTPT/diver/geop (1000hPA/500hPa)/U-V (925 hPA)/SPH) and few more , and honestly , getting one variable at a time ( in groups of 5 or less years), having to wait over 5hs to get it, its not acceptable. My TCC that I need to finish this year might not be ready because of this waiting time. So please , I am not the only one asking, we need answers, we need to know what is happenening and if it will ever improve downloading speed. I am waiting for 1 variable (T2M) since 2pm, and its almost 7pm here and yet, in progress.

Update:Recovering from connection error [(‘Connection aborted.’, RemoteDisconnected(‘Remote end closed connection without response’))], attemps 1 of 500
Retrying in 120 seconds

Really???Whats going on with this site?? my internet is working !!!

1 Like

Looks like the CDS system has had a big slowdown over the last couple days, and especially today.

From the live CDS system status page at Live — Climate Data Store just now, Thursday Oct 17 05:20 UTC, I see:

To me that looks like the completed request volume has been falling a lot over the last couple days, and especially today, with almost no data going out the last couple hours.

Given the typical queue depth, I’m guessing that’s due to the server side getting slower, not that the amount of client requests have dropped off.

Though maybe the metrics data for this dashboard is delayed, so the last couple hours or days always look low volume.

ECMWF CDS support folks, if you’re reading this:

  1. Do you know how up-to-date the info on that live system status page is? Is “0 requests last hour, and 15 requests last hour” a normal thing to see?
  2. It would be nice if the “Last n h” rows also included normalized requests and GB per hour values, to make it easier to compare speeds over time, and notice if there’s a big drop.
  3. Is the “requests” count here the number of requests which were issued, or completed, during that time period? I’m assuming completed. Might be nice to see the number issued, too.
  4. A count of “items” in addition to requests and GB might be nice too, since as I understand, the work done by the system, and thus request priority weighting and servicing time, depends more on that than total GB. Where “items” is (variables × levels × times, but not × grid cells).
  5. IMHO, would be nice to also see a graph of queue depth and average queue wait times, too.

I think that “0 requests last hour” might be real. ~30 minutes after I did that initial check, the 0 has rolled into the last 2 hours. Seems like the CDS might be pretty much stuck?

Hi all,
I downloaded ERA5 data in july using the old cds system and it was really faster.
Now I am struggling with this new version since one week and I had to re-run my script twice, because it got disconnected for many tries
I have this message at the request step and it lasts forever : 2024-11-20 09:28:31,242 INFO Request ID is 05cd7529-6b7e-46d9-b994-afca07484b0b
2024-11-20 09:28:31,274 INFO status has been updated to accepted

I am talking about the request and not the download step itself.

Same thing if you download monthly or daily data, no big changes.

Best