ZwiftPower update [November 2020]

Agh! That’s “new” from last year. Where we were able to see it weeks in advanced.
But good to know that it’s only 24 hours now, unfortunately that doesn’t allow desired much lead time to notify members that they’re signed up for the wrong category.

We checked in a ticket to extend that time to 5 days instead.


Great job reporting back!! Keep it up :+1:t3:


I’ll take it! TY!

Any updates @Flint_McInnis?

We don’t have any new updates right now, as we’re still working on these issues. We’re also making sure we are caught up after the U.S. holiday weekend here.

Thanks for your patience and for checking in with us. We’ll check back in with you when we have some more forward movement on some of these issues.

Ride On.


This will end in disaster. Zwiftpower which was made to solve a problem Zwift couldn’t (didn’t care about at all) solve is now under Zwift’s control.

Either 1. You think Zwift have changed their minds completly and that they have gone from wanting as many people paying as possible to caring about the racingcommunity or 2. You realise where this will end.

My question: Will you act when you see numbers which are suspicious with the same dedication as the guys on ZP or will it be more like Stravaleaderboards where you can laugh yourself the whole way to a KOM?


Hi Flint!
It’s now 3 days to our event and the sign ups are still not populating.

I’ve flagged this up with our Events team. Hope to get that solved for you promptly.

Hey Amy, thanks for checking in. At the moment, nothing has changed and it should still be ~48 hours before you see your events. The 5 day window is slated to take effect on one of our future rounds of maintenance as it requires code changes, testing, etc., to ensure that nothing breaks when we implement it. Sorry for any confusion and thanks for your patience!

Agh! I miss read that normal was 3 days prior, not 48 hours or less.

Hey @Flint_McInnis ,

Any status update on the current backlog for the FIT files in ZwiftPower?


Hey Josh,

I dropped a response over in the other thread you can check out.

The tl;dr is that we’re aware of it, working on it, and doing our best to get a solution that is robust and solid in an effort to make it a one-time fix.

Thanks for the response… appreciated!

ZwiftPower update Times are getting worse not better!
5 days to process fit files is really poor. Is something broken or is this how it runs now?

1 Like

@shooj: This is BS! You are a newbie software engineer who hasn’t dealt with real legacy systems. Something that is 5 years old is not legacy, even if it is running on a AS/400. Legacy hardware is another story, but my 10 year old xeon in my basement provides more HP than the AWS instances you are spinning up for ZP. You don’t break what is not broken, unless you want to break the entire system! I can already tell you combined functions into classes/libraries which combine read/output with backend processing, those items need to be separated. Also, it looks like you removed indexing which improved writes and processing on the backend. You improve the frontend speed by snapshotting the data. Yes, it increases storage requirements, but it also scales better.

1 Like

The scale of Zwiftpower is orders of magnitude smaller than many systems that process data in real time, such as Facebook, YouTube, Twitter and Instagram. These systems handle millions of requests per second, while Zwiftpower hasn’t been able to keep up with 1,000 fit files per hour since the migration to Zwift’s “robust, clustered server environment”.

So, while I understand you guys might have inherited a poorly designed system, the reality is that before the migration it was working fine for the end users, but it has become inexcusably useless lately.

At the very least, please amend the original November update by removing this:

1 Like

Hi @Flint_McInnis the link you posted seen a to be broken. I was interested in the details. Could you repost the link please?

1 Like

Ummmm… you took something that was working, and now have a Cleveland Steamer on the table. Blaming “legacy software” just seems like bad planning.