I seem to get these pretty much every day. I’m seeing a lot of slow responses today too, with each request taking 10+ seconds and this seems to be causing some knock on errors. Like I’ll tick an item to mark it as watched, it will take a long time to pop up saying that was succesful but then it still shows as unwatched.
We’ll be working on this more starting Monday. Sorry for all the random errors lately, we’ll get to the bottom of it and continuously roll out performance updates as we have them.
Experiencing quite a lot of Error 521 the last hours accompanied by low (to no) responsiveness while trying to mark items (as watched etc), which has been known before but this time in a disturbing amount, so that I feel the need to report it.
For a long time prior to the recent changes to the dashboard. It and nearly all watch history related pages on the site generated various Cloudflare timeout errors. 504, 512, 524, etc. This has gotten worse and worse to the point where it occurred greater than 9 times out of 10.
Since the change to the dashboard, it doesn’t time out any longer but the Up next to watch section almost never loads. No matter how long I leave it. Also, pages like Year in Review or Progress continue to time out far more often than they actually load.
Additionally, recently I will often find that items I know I have marked as watched don’t get saved as such. It is to the point where I’ve stopped trying to track movies watched and am only keeping up with TV.
This is clearly not an end device issue as it happens across multiple devices. And, I do recognize that I have a significant watch history, well over a 100k items. Is the service just not able to keep up with that volume?
Thanks
2 Likes
south america also, super slow website. before the dashboard load instantly and since a few weeks load like in 10 seconds, and the worst is that EACH section of the dashboard load with a 3 o 4 sec delay. like the “up next to watch” and “upcoming schedule”.
Also got a few random 524 errors or similar.
But why are those happening?
I recall that before it worked fine, including the profile page “next up to watch”, but now it takes ages to load if it ever does. If it do load, then the next page I try to load will most likely result in CF error 521.
Does anyone happen to know if the Plex agent is also affected by it and/or the API or just the frontend?
So I’ll update you guys as I figure out more of what’s happening, but as of right now, the root of this problem is that some bad API apps are stuffing user’s profiles with duplicate data. I just had to dedupe a user with over 2m plays. This kills database performance because the apps then repeatedly request the user’s history and jam everything up.
We’re currently working on a solution to find these bad applications and revoke their access until it’s fixed, as well as find existing user’s with bad data and squash their data.
I’ll update this thread when we’ve gotten something a bit more automated in place.
Thank you all for your patience… this has been bothering the hell out of us, so we 100% understand the frustration that must be happening for our users. 
3 Likes
well if your api doesnt limit per user updates, is also your problem. not some bad api apps.
every api that i ever used have a limit for most of the write/update requests…
so 2 millions updates on a single user is really bad api design.
How have things been today? We found the app and revoked access this earlier today.
We still need to clean up some user data, but things should be getting better.
Mine is very slow and still periodically failing. I have revoked access to every app on my account except ReelGood. But that is also the only one I have used in quite some time. Is it the problem?
Or does my account just need cleaning up?
On the dashboard:
- Up next to watch still fails to load more often than it works. But it seems to be working more often like 1 in 5 times instead of 1 in 50 like yesterday.
- Upcoming schedule and Recently Watched appear to be loading nearly every resfresh
Regarding the dashboard:
If my browser is zoomed too much when loading the dashboard, as soon as I scroll down to the “Up next to watch” section, it starts spamming GET requests to /dashboard/on_deck.
…or if the “Up next to watch” section is not visible for whatever reason when you load the dashboard (eg, when dev-tools is opened, docked to the bottom of the page & blocks the section), it also spams requests to the same endpoint.
Looking at the first 100 requests (it sends 500+), the server responds with a 200 status code every time and I can see the HTML in the response but it is not immediately rendered on the page. Sometimes the tab just locks up, sometimes after waiting 8-10 seconds it gets displayed properly on the dashboard.
I was able to reproduce it on Chrome and Firefox (Win10), after resetting browser data (at /settings/data), wiping my browser cache and signing out/in again, multiple times. Even on an alt account with barely any data, in incognito with no extensions.
I’ve no idea if this only happens to me, but could this be why others have been having issues loading the dashboard?
I can send a video/screenshots to staff if needed.
3 Likes
Wow… LiaraT, thanks for the direction to start looking. This doesn’t occur on mac browsers, so we never saw this during development.
Flipped on my windows box and it’s happening just like you said. We’ll let you guys know when this is updated.
1 Like
Hey Scott, I checked your account and it looks good. If the slowness is just on the dashboard, it’s probably the same problem as LiaraT is describing.
If not, I’m not sure. Is it a site wide slowness or specific pages?
Justin has a fix out for this. I had to clear my cache for trakt in order to get it to work again.
Big thanks again for finding this one!
2 Likes
Just the dashboard and its happening on Windows and Android.
Yes, it does seem to be better after clearing the cache
It seems harder to reproduce now, but still possible.
No matter how much I’m zoomed in, it doesn’t seem to happen again when I navigate to / refresh the dashboard while at the top of the page and scroll down “like a human”. Yesterday I was able to reproduce it those ways.
However it’s still happening when I navigate to the dashboard and immediately scroll down to ‘Up next to watch’ while zoomed in 150% or more in the browser.
Same happens when I remain on the dashboard, scroll down to ‘Up next to watch’ and then refresh the page (since the browser automatically moves the scrollbar quickly back to the position it was at before refreshing).
Reproduced a few minutes ago on multiple browsers with caching disabled & 150%+ zoom. 1080p display, 100% scaling in Windows settings - not sure if that matters.
It doesn’t happen lower than 150% zoom. The dashboard seems more responsive now too.
No problem Sean, happy to help.