[Openstack] dnsmasq bind error

If you see bind error while running openstack nova-compute or nova-network,

it probably is the generic dnsmasq launched by /etc/init.d

Just /etc/init.d/dnsmasq stop

and try again

[Openstack] rabbitmq rejects with invalid credentials

When rabbitmq log shows invalid credentials and rejects nova-services to join the cluster,

it probably means that the password is wrong.

nova-services in default use guest/guest (id/pw) for rabbitmq.

you can add (or change) the a user (or pw) with the following commands

add a user guest/guest

$ rabbitmqctl add_user guest guest

change the pw

$ rabbitmqctl change_password guest guest

[Openstack] Suggestion on the tight correlation to the mysql database and nove-service processes

If we stop the service, such as 'stop nova-network'
it would be appreciated if the corresponding service is deleted in mysql services table in nova.

[Openstack] nova.conf configuration

in nova.conf configuration flag, routing_source_ip flag should use 'numbered-ip' such as 192.168.1.1, but not the host name.
I thinkg openstack guys try to tell us by naming the flag as 'routing_source_IP'.
Since it is used as the input param to iptable-restore, thus if you use host name on 'routing_source_ip' nova-network will unexpected die.
If you try to run the nova-network again, it will be on deadlock, since the module that performs this use external lock in /var/log/nova/ and hangs in ps aux.
This makes you think nova-network is working fine but NO IT IS NOT.

It takes half a day to figure out what the heck is going on.

[OpenStack] possible self-deadlock?

Openstack manages external lock in /var/lock/nova directories.

If a service, (in my case, nova-network) dies while it is acquiring the lock, nova-network never be able to run properly since it is waiting for the lock that will be never released. (since the process that acquires the lock was terminated unexpectedly).

Openstack servies should release all locks with 'launch-all' command things like that.