AWS instance must not support NICS in-place update

Bug #1340065 reported by Pavlo Shchelokovskyy
6
This bug affects 1 person
Affects Status Importance Assigned to Milestone
OpenStack Heat
Invalid
Low
Pavlo Shchelokovskyy

Bug Description

According to AWS docs AWS::EC2::Instance supports updates to "NetworkInterfaces" and "SubnetId" properties only in UpdateReplace manner [1] but we have it accepting in-place updates in our AWS-compatible resource [2].

In light of ML discussion [3] our resource must be brought in accordance with AWS docs. As the change to AWS-compatible Heat resource was merged after Icehouse release, it involves simply reverting that change.

[1] http://docs.aws.amazon.com/AWSCloudFormation/latest/UserGuide/aws-properties-ec2-instance.html
[2] https://github.com/openstack/heat/blob/master/heat/engine/resources/instance.py#L229
[3] http://lists.openstack.org/pipermail/openstack-dev/2014-July/039429.html

Changed in heat:
assignee: nobody → Pavlo Shchelokovskyy (pshchelo)
description: updated
Revision history for this message
OpenStack Infra (hudson-openstack) wrote : Fix proposed to heat (master)

Fix proposed to branch: master
Review: https://review.openstack.org/105992

Changed in heat:
status: New → In Progress
Revision history for this message
OpenStack Infra (hudson-openstack) wrote : Change abandoned on heat (master)

Change abandoned by Pavlo Shchelokovskyy (<email address hidden>) on branch: master
Review: https://review.openstack.org/105992
Reason: After some thinking I reconsidered and tend to agree with huangtianhua - some update is better then none. The problem is that in order to align with AWS we need somehow first delete the old instance, create a similar one and stick it in the dependency tree at the same place. Needs more thought.

Revision history for this message
Pavlo Shchelokovskyy (pshchelo) wrote :

I am thinking about a following approach to tackle this discrepancy - if NICS (or Subnet?) are specified, in handle_update detach all the NICS and then raise UpdateReplace. The downside is that it makes 'update_allowed' for those properties effectively a noop, that might lead to some confusion (even with help strings for properties amended to describe this particular behavior).

Would such approach be acceptable?

Revision history for this message
OpenStack Infra (hudson-openstack) wrote : Fix proposed to heat (master)

Fix proposed to branch: master
Review: https://review.openstack.org/108002

Angus Salkeld (asalkeld)
Changed in heat:
importance: Undecided → Low
Changed in heat:
status: In Progress → Triaged
Revision history for this message
OpenStack Infra (hudson-openstack) wrote : Change abandoned on heat (master)

Change abandoned by Pavlo Shchelokovskyy (<email address hidden>) on branch: master
Review: https://review.openstack.org/108002
Reason: better to be better than AWS instead of strict consistency

Zane Bitter (zaneb)
Changed in heat:
status: Triaged → Invalid
To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.