Initial tests show no significant difference running disruptor e2e tests:

go test -tags e2e e2e/disruptors/pod_e2e_test.go

Before

ok  	command-line-arguments	137.136s

After

ok  	command-line-arguments	133.973s

This can be due to the disk not being a bottleneck in the test machine (which uses SSD)

Even if there's not a big improvement, it seems nicer on the environment (by means of less trashed SSDs :P)
It would be nice to have this by default, specially if we don't foresee etcd to grow big enough to eat the whole RAM.

How can I check if this is working?

I've tried running E2E_ETCD_RAMDISK=1 make e2e-disruptors, but I'm not seeing what I would expect. By describing the etcd pod in the cluster that gets created, I can see:

  volumes:
    - hostPath:
        path: /etc/kubernetes/pki/etcd
        type: DirectoryOrCreate
      name: etcd-certs
    - hostPath:
        path: /var/lib/etcd
        type: DirectoryOrCreate
      name: etcd-data

If I exec into the kindest/node, /var/lib/etcd is not mounted as tmpfs. From the code I guess path for the etcd-data volume should be /tmp/etcd instead. Am I doing something wrong?

Am I doing something wrong?

E2E_ETCD_RAMDISK=1 should be E2E_ETCD_RAMDISK=true or be omitted because it is the default.

@roobre You were right, I had a bug and wasn't passing the option from the e2e cluster config to the kind cluster config. Good catch.

I fixed it and checked that with the default settings the etcd pod mounts the etcd-data volume at /tmp/etcd and that the tmp directory is mounted on a tmpfs file system in the host node.