Jmeter testing with 5 replica
Here we are taking only 5 instance for each microservice and running the whole application. We are accessing all microservices by api-gateway only. We are using jetstream2 instance named as "Instance Viyad". It has CPU total 16 cores and 60 GB RAM.
These numbers were causing failure with replica 3, but able to handled with replica 5
- Expected throughput at each microservice
- Active threads over time
- Response Times Over Time
- Transactions Per Second
- Expected throughput at each microservice
- Active threads over time
- Response Times Over Time
- Transactions Per Second
- Thread Schedule
- Expected throughput at each microservice
- Active threads over time
- Response Times Over Time
- Transactions Per Second
-
Thread Schedule
-
Expected throughput at each microservice
-
Active threads over time
-
Response Times Over Time
-
Transactions Per Second
-
System’s capacity limits (at what load do you get significant failure rates). If we increase the number of users and requests per user and found out that system is failing at 6750 request that is 15 requests per user for 450 users. If we talk about spike testing, system started failing after 2000 threads.
-
Improvements that can demonstrate increased capacity.
- Why system continues to run even with injected failures. Right now system was able to handle such number of requests due to kubernetes loadbalancer feature with such number of replicas. But to support at large scale, we need to implement, auto scaling which can live other instance if no one is available or all dead.
- What is the impact on performance, compared to your previous load tests? We need to work to support more number of users, need to use message service so api-gateway does have much load and application can work much better.
User can get frustrated if lots of user access at the same time. As it does not support good number of users. It should show proper message while services are are down and redirect to customer support.