I'm testing the fix for that now. Also, this fix revealed an upstream bug in the management server. There is a path in advanceStart that can allow it to complete and return a null object without having successfully started the VM. The caller of this method (the orch service) doesn't care about the output, it relies on this method throwing an exception if the VM fails to start. The result is that calling startVirtualMachine can return a good result, but the VM never started.
In short I'll be filing a second bug soon. I'm testing a fix for this one as well. Unfortunately, master doesn't compile for me at the moment, so it's sort of difficult to test it against that and check them in to make cherry-pick requests. On Wed, Apr 17, 2013 at 11:37 AM, Chip Childers <chip.child...@sungard.com>wrote: > Hi all, > > Here's where we stand on ACS 4.1: > > One new blocker opened by Marcus today (which he assigned to himself): > > CLOUDSTACK-2067 > KVM virtual machines can fail to start but CS thinks they are running > > We also need to complete the release notes / upgrade instructions. > > AFAIK, that's it. So once we see 2067 get resolved, and we have the > release notes / upgrade instructions ironed out, we'll kick off a vote > for 4.1.0. > > -chip >
