Azure AD B2C throttling during performance testing

Question

Azure AD B2C throttling during performance testing

Parameswari Chandrasekar 0

Hi,

We have a pay as you go ad b2c account and trying to run performance tests in pre prod environments at 20 transactions per second. Azure starts throttling the request about 3 mins into the test run and they all start failing. Microsoft document states that there are no throttling issues with a pay as you go account and we should be able to ramp up performance tests to scale. Any suggestions on how to get past this?

Locked Question. You can vote on whether it's helpful, but you can't add comments or replies or follow the question.

Christos Panagiotidis 3,551 Reputation points

2026-07-29T10:34:07.5066667+00:00

Pay-as-you-go billing does not remove Azure AD B2C protection limits. The documented limits include 200 requests per second per tenant and 6,000 requests per five minutes from one IP. A single sign-in transaction consumes multiple endpoint requests—typically four before optional MFA or federation requests—so 20 user transactions per second can produce substantially more than 20 B2C requests per second. A test originating from one load-engine IP can also reach the five-minute IP limit after several minutes.

Calculate the expected request consumption for the exact user flow, then inspect 429 responses and their retry information. Reduce the ramp rate, avoid repeatedly downloading discovery metadata or keys, and apply exponential backoff. Use realistic test distribution rather than attempting to bypass protection controls, and do not run disruptive tests against production. If the calculated production requirement genuinely exceeds the published limits, open an Azure Support request with the tenant ID, policy name, expected peak rate, flow features, source distribution, and correlation IDs to request a reviewed service-limit increase.

Christos Panagiotidis 3,551 Reputation points

2026-07-29T10:34:07.5066667+00:00

Pay-as-you-go billing does not remove Azure AD B2C protection limits. The documented limits include 200 requests per second per tenant and 6,000 requests per five minutes from one IP. A single sign-in transaction consumes multiple endpoint requests—typically four before optional MFA or federation requests—so 20 user transactions per second can produce substantially more than 20 B2C requests per second. A test originating from one load-engine IP can also reach the five-minute IP limit after several minutes.

Calculate the expected request consumption for the exact user flow, then inspect 429 responses and their retry information. Reduce the ramp rate, avoid repeatedly downloading discovery metadata or keys, and apply exponential backoff. Use realistic test distribution rather than attempting to bypass protection controls, and do not run disruptive tests against production. If the calculated production requirement genuinely exceeds the published limits, open an Azure Support request with the tenant ID, policy name, expected peak rate, flow features, source distribution, and correlation IDs to request a reviewed service-limit increase.