Duplicated data in request

[yatoogamii]

I try to get anime data with postman using this request : https://api.jikan.moe/v4/anime?q=shingeki+no+Kyojin&type=tv&min_score=8 who give me successfully 6 differents anime:

• Shingeki no Kyojin
• Shingeki no Kyojin Season 3
• Shingeki no Kyojin Season 2
• Shingeki no Kyojin: The Final Season
• Shingeki no Kyojin: The Final Season Part 2
• Shingeki no Kyojin Season 3 Part 2

Now If i change the request to not get all the anime but only one by one using pagination (with page and limit args)

https://api.jikan.moe/v4/anime?type=tv&min_score=8&q=shingeki+no+Kyojin&limit=1&page=1

• Shingeki no Kyojin

https://api.jikan.moe/v4/anime?type=tv&min_score=8&q=shingeki+no+Kyojin&limit=1&page=2

• Shingeki no Kyojin: The Final Season

https://api.jikan.moe/v4/anime?type=tv&min_score=8&q=shingeki+no+Kyojin&limit=1&page=3

• Shingeki no Kyojin Season 2

https://api.jikan.moe/v4/anime?type=tv&min_score=8&q=shingeki+no+Kyojin&limit=1&page=4

• Shingeki no Kyojin: The Final Season

https://api.jikan.moe/v4/anime?type=tv&min_score=8&q=shingeki+no+Kyojin&limit=1&page=5

• Shingeki no Kyojin: The Final Season Part 2

https://api.jikan.moe/v4/anime?type=tv&min_score=8&q=shingeki+no+Kyojin&limit=1&page=6

• Shingeki no Kyojin: The Final Season Part 2

(The order and the item duplicated can be different sometimes)

As we can see there is duplicate data and I think it's not normal ? Unless I didn't understand something.

Can you help me ?

[irfan-dahir]

Sorry for the late response. I think this is a caching error on the load balancer given the data is being coming from multiple instances of Jikan. But I don't understand why the ordering would be different. 🤔

I will take a detailed look into this soon. Good find!

[zakaoai]

Currently : https://api.jikan.moe/v4/anime?q=sao&sfw
Return 2 times id 55994. Do we need to manually filter duplicate because 1.5 year later we still got this issue

[pushrbx]

I've triaged this the best I could: I couldn't reproduce this behaviour locally and my in own self-hosted instance. This is a cache/infrastrucutre issue I believe. In case of the public api there are multiple workers, and there can be race conditions, where two worker is trying to scrape the same anime/manga, and inserts them into mongodb. There is no locking mechanism to sync the workers. I'm going to create an issue for this and reference all the similar ones under it.