Closed lukeoguro closed 1 year ago
I'm still getting that address not found error when running a test:
--> edamame run --file ./k6_tests/test1.js 18:54:52 羽22m 30.212s
[06:55:42:378] ℹ Reading test script...
[06:55:42:381] ✔ Successfully read test script.
[06:55:42:381] ℹ Initializing load test with 1 load generator (5000 VUs per pod)...
[06:55:43:997] ✖ Error running test: Error: getaddrinfo ENOTFOUND a90b4e77762a244709151e2cbadc4523-982894087.us-west-2.elb.amazonaws.com
This is a new cluster, and I've switched to the m5.xlarge
type to avoid conflicts with availability zones
The test appears to work if I try to run a second time, so I think this is just a timing issue for the external service.
Thanks for testing Ginni! I believe this stems from running tests too quickly after edamame init
— we might want to add a polling measure to make sure the DB API service is ready before saying "welcome to Edamame".
Yeah I figured it might other than the changed code here.
I took a quick testing break to eat, but I'll get back on it here in a few and if everything is above board will get things merged! 🙌
Everything looks good to me! Nice work!!
This ended up being a bit of an overloaded PR, but the main change is to:
--path
flag to--file
foredamame run
so its closer to thekubectl
convention--vus-per-pod/-v
flag to customize the number of VUs per pod.To test this, run a test script with and without the flag.
kubectl get nodes
andkubectl get pods -o wide
should show you how the number of nodes will change with the flag. After the test finishes, make sure the nodes are scaled back down to 0.The other major change is I've performance-tuned
statsite
to handle up to 2 million metrics a second. There may be some issues with the availability ofm5zn.xlarge
instances, which in that case, you can manually change the value tom5.xlarge
in thestatsite_nodegroup_template.yaml
file.Resolves #97; Resolves #36; Resolves #90