The Movie Database Support
Support → API Support
We have recently gone public with the new MB3 server and classic client in beta and are experiencing a huge amount of time-outs trying to access the tmdb API. Has something changed recently that we missed?
Reply by Travis Bell
on May 20, 2013 at 10:17 PM
Nope, everything has been running very smoothly actually.
We had some invalid JSON being served for about 3 hours on Friday night (May 17th). Other than that no reported or known issues.
Does each end user connect to our service on their own or do they connect to your servers and you run updates to sync things up so there's only the one "client" as far as TMDb is concerned? I only ask because of the rate limiting, which is imposed on a per IP basis.
Reply by ebr
on May 20, 2013 at 10:24 PM
No, they all connect on their own - same api key though. It's odd because we've had no real issue with MB2 nor to this point with MB3. But there are about 1000 extra servers out there now.
We've been bumping down our requests to single threaded even and still seem to be hitting time outs hard from a number of clients.
Reply by Travis Bell
on May 20, 2013 at 10:30 PM
When you say timeout, is it a legitimate HTTP timeout or are you getting an error code back? We serve back a 503 error when the rate limiting is hit.
We haven't made any changes to the infrastructure though.
Reply by ebr
on May 20, 2013 at 10:41 PM
We are getting http timeouts I'm pretty sure but I'll have to dig through some user logs and see if we can see the actual response code.
Reply by Travis Bell
on May 21, 2013 at 10:07 AM
One other thing I just thought of, does MediaBrowser happen to do an hourly scan of a users library by any chance? Say, at the top of every hour?
Reply by ebr
on May 21, 2013 at 12:37 PM
I believe it is set to every two hours by default but that won't hit y'all unless there is something new.
Reply by Travis Bell
on May 21, 2013 at 1:33 PM
Ok, I only ask because there's an app out there (haven't looked that much into it yet) hitting our API every hour. It's a massive spike pushing the API to over 3,000 r/s for a brief period. If it's Media Browser, which I might be inclined to think it is, you have ~thousands(?) of users hitting the API at once and it's pretty much a recipe for disaster… while our system can handle this now, what happens when the userbase doubles? Whoever this is, I wanted to talk to them about randomizing the scan time per install so it's not an issue.
Reply by ebr
on May 22, 2013 at 3:04 PM
I don't think that is us but let me know if you find out differently. We cache the information locally and only hit you guys if something has changed and we don't have it locally and this would never be a big spike on any kind of regular interval. I'm sure we've generated a lot of traffic in the past week with the public beta of MB3 but it wouldn't be anything like you are describing.
Reply by Travis Bell
on June 10, 2013 at 10:46 AM
Hey,
Did you ever dig into this and find out anything more?
Reply by ebr
on June 10, 2013 at 11:02 AM
Yeah, it was not a limit refusal it was an actual timeout and it was something related to the caching that was going on. Nothing on your end so this issue is resolved.
Did you get my email last week re:donations?
Reply by Travis Bell
on June 10, 2013 at 11:38 AM
Cool. Good to know.
Don't think so. What email did you send it to? travis [at] themoviedb.org is best.
Reply by ebr
on June 10, 2013 at 12:14 PM
I used the webform on here. I'll send you one directly. Thanks.