A few comments on some troubles I have had during my test deployment..
following https://help.ubuntu.com/community/UEC/CDInstall
Using the 10.04 beta install disk, installing all components on a single piece of hardware
First of all, when trying to run either emi-DF89105E or emi-DF89105E, which are the only 32 and 64 bit ubuntu server images currently available in the store, ec2-init hangs for about an hour before the instance starts up sshd, which is damn annoying, and I haven't figured out how to get past that on first run (the first thing I normally to when I can log in is apt-get remove ec2-init).
Once ssh is running, I have found that the user I need to use is actually "root", not "ubuntu" as described in the documentation. Even then, pubkey authentication occasionally will fail on some instances.
One thing I have also had trouble with is my instances not surviving reboots. The files for the instances usually seem to be there after a reboot, and euca-describe-instances says they are pending for a while, and then terminates them and wipes all the files (including the console log). I figured out that I could set MANUAL_INSTANCES_CLEANUP to 1 in the eucalyptus.conf, which saves the images from being wiped (sometimes), but even then, I have no idea how to launch the instance again. I even used ps aux to grab the kvm command and all the proper arguments, but running the command manually still failed.
How am I supposed to install kernel updates etc if all my instances die when I reboot the host? Is there some sort of euca-save-instance euca-restore-instance command that I am just not seeing or something? I am not really even sure what logs I should be looking at. The only ones I have really been looking into was the cc.log, nc.log, and the console.log in the instance directory. Any tip or advice on something I may have overlooked would be greatly appreciated, thanks.
Bookmarks