Closed nickumia-reisys closed 2 years ago
I suspect that the test failed because of this,
It was trying to start the ebs-app and just hung there, so it would make sense..
This is the best screenshot that I have for proof that the issue was with there not being enough nodes to run 2048 and ebs test workloads on,
Could that have something to do with setting '--max-pods=20'
? That would mean only 20 pods max would be scheduled on the single node, even if it had capacity to schedule more.
(I don't know what the default is when you don't specify that argument.)
That could definitely have an effect! @mogul I could leave that option off. I guess I hadn't thought about that haha..
(I don't know what the default is when you don't specify that argument.)
That's also true. I mean, we could specify a really high number and then let the CPU and Memory determine if a pod can run.
It looks like the default varies by instance type, and they're listed here.
Related to https://github.com/GSA/data.gov/issues/3668