On 09/07/2016 05:25 PM, Dmitry Tantsur wrote:
Hi all!Today while playing with my installation I noticed that we do try to run cleaning for nodes in maintenance mode. This leads to a somewhat confusing result, because we no-op heartbeats for such nodes. So cleaning gets stuck in "clean wait" forever [1]. However, it seems like some folks find it a convenient feature. This way they can ask Ironic to boot the ramdisk and make it wait for operator's commands. It's a fair use case, but I still find the current situation confusing. We've ended up with these few options: 1. Ensure we don't run cleaning for nodes in maintenance mode. I've proposed a patch [2] banning most of provision verbs from working in maintenance. However, we still want to allow deleting an instance, which still results in cleaning. 2. On receiving a heartbeat in {CLEAN,DEPLOY}WAIT and maintenance on [3], move the node to {CLEAN,DEPLOY}FAIL (optionally without powering it off, so that IPA stays running).
It looks like this approach does not cause contention, so I'll look into it first. I will keep patch for option #1 around as well, in case we want to explore it.
3. Document this as a desired feature and don't change anything. What do you think? [1] https://bugs.launchpad.net/ironic/+bug/1621006 [2] https://review.openstack.org/#/c/366793/ [3] https://github.com/openstack/ironic/blob/master/ironic/drivers/modules/agent_base_vendor.py#L474-L478. __________________________________________________________________________ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
__________________________________________________________________________ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
