Nova compute service does not shutdown gracefully when it is terminated abruptly during create server api

asked 2014-02-12 04:29:05 -0600

Aswad Rangnekar gravatar image

updated 2014-02-14 18:58:14 -0600

smaffulli gravatar image

Hi,

Ref Nova graceful-shutdown BP: https://blueprints.launchpad.net/nova...

I am testing Nova graceful-shutdown BP for "create server scenario" on "single node Openstack installation using Devstack".

When I terminate the nova-compute processes, it does not wait for run_instance() from nova compute manager to complete. The processes terminated abruptly. According to the BP; nova compute services should wait for long running calls and it currently supported SIGINT and SIGTERM processes kill signals. Where am I going wrong while testing this scenario? Is create server API call supported by this BP ?

I performed following steps for testing:

  1. Insert pdb/time.sleep(20) in nova.compute.manager run_instance()
  2. Launch a new instance using dashboard (nova compute processes will stop at the point where pdb is inserted or time.sleep is introduced as described in step 1.
  3. Kill nova compute processes using any one SIGINT or SIGTERM signal as shown below
$ sudo kill -2 <nova_compute-processes-id>
OR
$ sudo kill -15 <nova_compute-processes-id>

Is there any alternate way to test this blueprint?

edit retag flag offensive close merge delete