In the following post, we are going to talk on how to perform a set of benchmark tests on the ETCD cluster of Openshift 4 with root disk over SAN.
Prerequisites
Using the etcdctl check perf
tool:
Step 1. Testing the ETCD
Before performing any ETCD benchmark testings please collect oc must-gather logs.
Step 2. The benchmark threshold results set by the upstream community
Measure |
Threshold limit |
Throughput |
< 150 * 0.9 writes/s (s) ; < 1000 * 0.9 writes/s (m) ; < 8000 * 0.9 writes/s (l) ; < 15000 * 0.9 writes/s (xl) |
Slowest request |
> 500 ms |
Standard deviation |
> 100 ms |
Step 3. Testing the ETCD with kube-burner
In this point we are going to extend the benchmark tool described on the Step 1 for the ETCD cluster on OCPv4.X.
Before performing any ETCD benchmark testings please collect oc must-gather logs.
Run the must-gather through the etcd.sh:
At this point we will need to tune the workload file configmap-scale.yaml
as described below:
Running the kube-kurner
service in a dedicated terminal:
Open a new terminal and run the following commands:
Once those steps are finished and all the data has been collected, try to perform a oc must-gather logs collection.
Run the must-gather through the etcd.sh:
For more informations on the etcd.sh.
Run the fio_suite: