Hi, I was trying to sort out a ulimit (open files) issue with a small Hadoop cluster, and some of the remedies I came across on the web involved tweaking the pam configuration and rebooting. I've been using whirr for a while now in EC2, and I am currently using version 0.7.0 to stand up a CDH-based cluster using 64 bit Ubuntu 10.04 images.
When I rebooted my nodes, they came back up (i.e. AWS reports them as running and healthy, using the basic cloud watch monitoring), but I can no longer SSH to them using the ec2-user credentials/keys that normally work in a whirr-instantiated cluster. This applies to every node in the cluster, and I've re-launched the cluster several times now -- 100% reproducible. Is this a known limitation? Does anyone know of a customization to the installation/configuration function scripts that will ensure that the ssh config persists through a reboot? thanks, Evan
