Instance fail to run

[Etienne Goyer]

On karmic running Eucalyptus 1.6~bzr672-0ubuntu3, instances fail to start. I have not been able to track down the cause. I systematically get the following:

cloudmaster@uec-frontend:~$ euca-run-instances -t c1.medium -k warthogs-key emi-073A1120
RESERVATION r-5254095A admin admin-default
INSTANCE i-430707D2 emi-073A1120 0.0.0.0 0.0.0.0 pending warthogs-key 2009-09-14T23:52:45.132Z eki-F2801528 eri-50131678
cloudmaster@uec-frontend:~$ euca-describe-instances
RESERVATION r-5254095A admin default
INSTANCE i-430707D2 emi-073A1120 0.0.0.0 0.0.0.0 pending warthogs-key 0 c1.medium 2009-09-14T23:52:45.132Z warthogs eki-F2801528 eri-50131678
cloudmaster@uec-frontend:~$ euca-describe-instances
RESERVATION r-5254095A admin default
INSTANCE i-430707D2 emi-073A1120 0.0.0.0 0.0.0.0 terminated warthogs-key 0 c1.medium 2009-09-14T23:52:45.132Z warthogs eki-F2801528 eri-50131678

Attached nc.log from the only node controller. There are a bunch of error in there, but I am not sure which are relevant and which are not.

[Soren Hansen (soren)]

Etienne, could you check the the node controller is running as the eucalyptus user and that the eucalyptus user is a member of the libvirtd group? Also, does /var/run/eucalyptus exist on the node controller?

[Etienne Goyer (etienne-goyer-outlands)]

Unfortunately, I had to revert back to 1.5 for a specific project, and I do not have the hardware to run two clusters in parallel (a stable one on jaunty/1.5, and one for testing on karmic/1.6). I will not be able to test any earlier than late next week. Sorry! :(

[Daniel Nurmi (nurmi)]

I believe that there are two problems here; first, the latest apparmor/libvirt is preventing the NC from being able to start VMs:

https://bugs.launchpad.net/ubuntu/+source/eucalyptus/+bug/431090

second, the 'nc-stat' error is caused by /var/run/eucalyptus directories are not getting the right ownership upon reboot of the machine(s):

https://bugs.launchpad.net/ubuntu/+source/eucalyptus/+bug/431114

This problem should not have a dire effect on the NC, but will effect the CCs ability to run in MANAGED mode.

[Thierry Carrez (ttx)]

Both blocking bugs have been fixed, so we'll suppose it's fixed now. Please reopen if you can still reproduce it. I will conduct some testing on instance-running soon and will reopen/file new bugs if any issue remains.