9 months ago
Hi everyone
I have an application on Railway, and it has started to scale well.
In order to ensure that I won't have problems with the upcoming increase in demand, I'm running a test using Artillery to validate one of my APIs.
I would like to confirm if there is a rate limit on Railway, and if it is possible to adjust it?
I need to perform a load test of 5,000 to 10,000 requests per minute.
I started this test and was blocked:
Timeout awaiting 'request' for 10000ms
0 Replies
9 months ago
@Brody can confirm, but I don't believe there is a rate limit
I had to change my connection to get access again.
My main IP (home) is blocked from accessing any service I have on Railway
im using my 5g connection rn
if i try from the connection i used to make the load test, i get this: ERRCONNECTIONTIMED_OUT
it just released my IP
I believe I reached the rate limit and was blocked for a few minutes
https://docs.railway.com/reference/public-networking#domain-rate-limits
I found this information, but I don't think I achieved it.
9 months ago
yes there are rate limits, they are per ip though, so if you hit your service hard from a single ip you will get rate limited, thus you are unlikely to hit a rate limit with genuine traffic
9 months ago
i think there would have to be something critically wrong with your application's design for you to hit these limits with legitimate distributed traffic
I would just like to confirm if I understood correctly, my IP was blocked because it exceeded the limit of 100 concurrent connections, correct?
in the load test I performed, I tried to make 120 requests per second, and the process takes about 300ms, so it may have exceeded the 100 simultaneous connections
9 months ago
correct
I will confirm, but this can happen in the current model (as bad as it may seem)
9 months ago
it happened because all the traffic came from a single ip, you could do far more traffic if you didnt make all the requests from a singular ip
Yes, I understand that
The point is that the requests come from a client (same IP) to here, currently around 500 requests per minute.
I made several improvements to this service to meet the promised demand, I don't believe it will reach 100 per second, but if it does happen I would like to know what I can do to adjust it.
9 months ago
thats not something that can be adjusted by the end user
@Brody if i share the static IP of my client, is it possible to set like in a white list?
9 months ago
thats also not something that can be adjusted by the end user
9 months ago
sorry no, limits are there for a reason, and we have yet to see a single user hit them with a correctly behaving application
9 months ago
^
9 months ago
no its not an issue with what your application does, its the amount of requests
Obviously these are peaks, it's not 24 hours a day
but that's okay, if it's not possible, I'll try to keep it up as long as I can, if necessary I'll try another solution
9 months ago
happy to help!
9 months ago
!s
Status changed to Solved brody • 9 months ago