Is there a queue penalty for failed requests?

Datasets and Usage C3S - Datasets and Usage
1

Is there a queue position penalty for failed requests on subsequent queued requests?

And tack on question, regardless of failures. I understand that there are penalties for previous heavy requests. How long before such penalties reset? How long should I wait until making a new request if I think that a previous request has imposed a queue penalty on my request?

Context:

When looping through CERRA height level requests, each request was relatively uniform for queue time (5-20min after previous request finished) and runtime (5-40min) until a request failed. After the failure (Failed to write chunked buffer 2225137/2286577 (Illegal seek)), all lined up requests in my queue remained as “Accepted” for 4 hours before I aborted.

Each request is for 5 vars, full month, combined analysis and forecast 1, 2 hour lead times. Repeating requests for each month. Trying to get historic data for 30 years. This seems like the the most efficient request structure based on documentation.

2

Hello Dan,
Trying to download CERRA pressure levels here with almost the exact same request, i am experiencing huge waiting time per request since monday approximately 5pm. Does it coincide ? I was thinking about same thing as you about a potential penalty for constant requests but tried to pause my requesting loop and it didnt change a thing. can you confirm me you are still experiencing the issue?

3

I am still experiencing slow requests. Its seems to be a technical issue with MARS. See response from ECMWF below. The live status is now also showing System Down.

"There are some technical issues on MARS which are unfortunately significantly impacting CDS users attempting to retrieve data stored on MARS tapes e.g. CERRA data. The Data Stores administrators and the MARS team are actively working at resolving the problem as soon as possible.

Please be aware that there is a risk that some requests will have to be cancelled by the administrators."