Axuda de The Movie Database
Axuda → Axuda da API
Changes in API made it slower?
Publicado por CouchPotatoSTAFFMOD o 22 de xullo do 2014 ás 3:40PM
The last couple of days I been having some connection issues. I haven't touched the code that caches stuff from the api for more then a month, but looking at the API stats I see this: http://cl.ly/image/1P1W3D3D3H1e
Can someone help me out? Searching sometimes returns instantly, but usually takes 1.5+ seconds. Sometimes even 5s. If I had a simple app that wouldn't be an issue, but I have around 150 req/s hitting my server. If data isn't returned fast enough, the connection is in use and new users can't connect.
I cache both search results and movie details for 24h. The 400k requests is under 5 req/s on your server, but that tripled in the last couple of days.
Like I said, a few days ago it worked fine.
Resposta de Travis Bell
no 22 de xullo do 2014 ás 6:28PM
Nope, no known issues. Latency and average response times have stayed flat for about as long as my graphs go back.
The dip in stats was a problem that only affected our stats servers. This was mentioned here: https://plus.google.com/105811688148229485626/posts/3bBRKTG3Pbk
Are you only noticing this from a single server? And also, on only search or all of the API?
Resposta de CouchPotato
no 23 de xullo do 2014 ás 3:36AM
I've disabled the search, see if that slows it down. The other calls have the same issue (getting info takes some time, including Credits and Alternative Titles, which are cached for 3 days).
I have this issue both locally (located in The Netherlands) and from the server (located in the US, east-coast)
Keep you posted.
Resposta de Travis Bell
no 23 de xullo do 2014 ás 10:16AM
One thing I’m wondering is whether or not you might be hitting our rate limits... do you have an idea of the cache hit ratio of the requests hitting your server? Our rate limits are based per IP so in a proxy situation like your setup, it’s not that hard to imagine you might start to be up against the wall (especially since the growth on your end has been consistent). We rate limit at 30 requests per 10 seconds (but there’s some weirdness here which is why I posted this a while back--although it has yet to be pushed into production).
I’m obviously not sure how things are setup on your end but is it possible for you to pass some of the calls straight to us rather than go through your server? This would make every end user’s IP the tracked IP and free up resources on your end.