nova compute does nothing on receiving sighup signal

[Abhishek Kekane]

Nova compute service does not listen to sighup signal actually it should reload the configuration file.

Steps to reproduce:
1. Run nova compute service as a daemon
2. Send sighup signal "kill -1 $parent_id_of_compute_service" or "kill -HUP $parent_id_of_compute_service"

Whereas if you send sighup signal to nova-api or nova-conductor then it reloads the configuration files.

[Sylvain Bauza (sylvain-bauza)]

The nova-compute service supports SIGHUPs and you should see a log saying it got the SIGHUP signal if you do the above :
https://github.com/openstack/nova/blob/088521591ad2ca20d8267d53d72aaca4441675d6/nova/compute/manager.py#L540

That said, reloading configuration files on SIGHUP is currently not fully supported for all configuration options in Nova. Most of the effort was tracked by https://docs.openstack.org/newton/config-reference/mutable.html

At the moment, we only have a very few mutable config options:
 - [libvirt]/live_migration_completion_timeout
 - [libvirt]/live_migration_progress_timeout
 - debug (inherited from oslo.log)
 - log_config_append (inherited from oslo.log too)

Could you please verify the above config options and check those can be changed without requiring a service restart ?

I'm closing the bug now as I think the expected behaviour is correct but feel free to reopen the bug if you see those 4 options not being mutable.