NuGet download stats have been broken for several weeks

[Aaronontheweb]

We track downloads of our packages every week as just a general health metric for our project - we noticed that the numbers dropped several standard deviations below their historical averages some time after 7/21/2025

We're not the only ones who've noticed:

• https://www.reddit.com/r/dotnet/comments/1mqsfp6/nuget_reported_download_stats_dropped_sharply_in/
• https://x.com/NuGetTrends/status/1942042374294610260

Is this an issue on NuGet's end?

[agr]

We are aware of the issue. Logs from one of our CDN infrastructures are not being processed, we're investigating why. Once the issue is mitigated and queued logs processed, we expect to have download data backfilled since the incident start.

[agr]

The underlying issue is mitigated, but it will take a while to process the backlog accumulated. We'll post an update when it is done.

[Aaronontheweb]

Thanks @agr

[jodydonetti]

Thanks @agr !

[HofmeisterAn]

@agr Thanks for looking into it. If I understand correctly, statistics should have been working again since last week, but the missing data from before still needs to be imported. From my side though, it looks like the statistics aren't updating at all anymore, which seems even "worse" than before.

[agr]

We had more issues over the weekend, all mitigated now. We're still catching up on missed logs. In the last week we caught up about 16 days and are still 8 days behind, so it'll take several more days to process all backlog.

[agr]

An update about recovery process: we caught up with daily routine log ingestion, but to speed that process up with had to "stash" some unprocessed logs away (it contains downloads from end of July to around August 10th). We're now slowly feeding that stash back into log processing pipeline.

[jodydonetti]

Thanks @agr , something seemed still off, now it makes sense.

[agr]

Apologies for the late update. As of October 2nd, we finished catching up with log ingestion. We ended up losing about 4 days of download information. Changes were made to the log ingestion pipeline to harden it against similar issues and alerting rules to make any other pipeline failures more visible.

[jodydonetti]

Thanks Andrei, appreciate the update.