Timeout on https://dataloen-api.bluegarden.dk/api/input/salary/XXXXX/forms Endpoint
Hi Support Team,
we do currently investigate an error when processing payroll to Dataloen for a mutual customer. When investigating the issues they reported we found some Timeouts (cancelled calls) to the API after 1.7 min. And we would like to check with you what could be their case.
Example:
- 10/23/2023, 2:49:27.564111 PM BST
- Connection ID ad70d67a-5755-466d-bb83-3440a7d30955
- https://dataloen-api.bluegarden.dk/api/input/salary/11708099/forms ?from=2023-09-15&to=2023-10-14&page=11&state=Committed fails
- https://dataloen-api.bluegarden.dk/api/input/salary/11708099/forms?from=2023-09-15&to= 2023-10-14&page=50&state=Committed is successful
-
Hi,
It looks like its a performance issue on our end. The 1.7 min is 100 seconds, which I think is the default timeout value. The customer in that example is by far our largest customer, and we routinely have issues with performance for them in different parts of our system.
I think maybe you can reduce the response time by making fewer requests with increased page size (using querystring pagesize=X). Right now you are using the default page size of 20, which means we get a lot of requests at the same time when you want to fetch all the data. This causes the response time to gets increasingly worse for each request, as you can see from this example from our request log:
Each request has to fetch the all the data every time, and the data is filtered afterwards, according to the pagination. So I think fewer requests with larger data chunks should stress our server less.
Either way, sorry about the poor performance, but I don't think we'll get time to fix that in the near future. -
Im not sure, this is mostly guesswork. 200 perhaps?
Another idea: I assume you send all the requests simultaneously, since you probably just want all the data for the given period. Maybe adding some chunking to the requests could also help, such that you only send 5-10 at a time, and wait for them to complete before sending the next chunk.
Log ind for at efterlade en kommentar.
Kommentarer
3 kommentarer