Merge lp:~sberg-l/charms/trusty/nova-compute/next into lp:~openstack-charmers-archive/charms/trusty/nova-compute/next
| Status: | Superseded |
|---|---|
| Proposed branch: | lp:~sberg-l/charms/trusty/nova-compute/next |
| Merge into: | lp:~openstack-charmers-archive/charms/trusty/nova-compute/next |
| Diff against target: |
69 lines (+18/-1) 6 files modified
config.yaml (+4/-0) hooks/nova_compute_context.py (+4/-0) revision (+1/-1) templates/icehouse/nova.conf (+3/-0) templates/juno/nova.conf (+3/-0) templates/kilo/nova.conf (+3/-0) |
| To merge this branch: | bzr merge lp:~sberg-l/charms/trusty/nova-compute/next |
| Related bugs: |
| Reviewer | Review Type | Date Requested | Status |
|---|---|---|---|
| Billy Olsen | 2015-08-21 | Needs Fixing on 2015-08-31 | |
|
Review via email:
|
|||
This proposal has been superseded by a proposal from 2015-09-02.
Description of the Change
Add a configuration option for nfs-mount-options. This is necessary for nova to work after configuring cinder to use Tintri as a backend.
charm_unit_test #7821 nova-compute-next for sberg-l mp268680
UNIT OK: passed
charm_amulet_test #5934 nova-compute-next for sberg-l mp268680
AMULET OK: passed
Build: http://
| Ryan Beisner (1chb1n) wrote : | # |
Thank you for your work on this.
Since we had been talking in #juju, I wanted to chime in with a few thoughts to help make your project reviews go as smoothly as possible.
I think that my colleagues will like to see and review this in a batch along side any other proposed changes to other charms, so that we can understand and review the changes in full context.
Until a more official review, some things to double-check as potential blockers:
# Config
While I've not taken a deep dive into this, I know that (generally speaking) many config options have been relocated or deprecated, depending on which OpenStack version is in use. So I'd recommend double-checking all of the relevant config options and template placement against the upstream docs, such as:
http://
http://
http://
A couple of questions that jumped out at me after having a quick peek:
Will you also need to enable a volume_driver for nfs in nova.conf?
Will you also need to specify a nfs_mount_
# Supported Releases
Take note that with each OpenStack charm, we develop and test with the expectation that each charm will succeed on all currently-supported releases. Currently, that is Icehouse, Juno, Kilo, with Liberty up next in active development. As a handy reference, this page has a chart that shows the supported release timeline:
https:/
# Contacts
As you know, you can generally reach one of us in #juju. For a-sync questions, the juju mailing list is also effective in drawing input/feedback with regard to general design questions or issues. I highly recommend joining that, as it's fairly low-traffic, and definitely has the right audience.
https:/
Thanks again!
| Skyler Berg (sberg-l) wrote : | # |
Thanks for the feedback, Ryan!
nfs_mount_options was added to the [libvirt] section of nova.conf in Icehouse and has continued to exist in the all subsequent releases. As such, I added to the templates for all Icehouse, Juno, and Kilo.
I do not need any additional options other than nfs_mount_options.
As for other changes, I do not anticipate any. This change should be isolated. Soon I will be uploading my cinder-tintri charm. In that charm's README I will instruct users to set this option appropriately. However, the code in my charm will not touch nova-compute at all.
| Billy Olsen (billy-olsen) wrote : | # |
Ah this is where you were going with the general config options? From our discussions on IRC, I understood you wanted general config settings set rather than ultimately seeing it this way.
Although, this is certainly one option of solving it, I think you are probably better going down the path of creating a subordinate charm and adding a tintri relation interface. As a simple subordinate charm, it could easily relate to both the nova-compute charm (which would react to tintri-relation-* hooks) and to cinder (also reacting to tintri-relation-* hooks).
For example, let's go with the idea of a tintri-storage charm which is a subordinate. The subordinate means it will be colocated with each unit it is paired with. In this charm, there would be the options for connecting to the tintri storage device (e.g. username, password, ip address, etc). It would provide the tintri interface, which shares access information to the related service (e.g. on the relation it would share the ip, password, etc). Then the nova-compute charm would be augmented to understand the tintri interface. On the tintri-changed-* hook it would read the relation values and configure the local storage access and any nova-compute parameters needed (ideally a new TintriContext or NFSContext etc would be created here), which would also include the NFS mount options etc.
I'm going to mark this as needs fixing as I think you really (ultimately) want what I've outlined here. That is to give the user one place to configure the tintri settings for the environment, and then let the charm relations handle the rest for you.
Ping me (wolsen) on irc if you want to discuss this any further or have questions. I'll be more than happy to help out!
| Billy Olsen (billy-olsen) wrote : | # |
btw, when it is ready for re-review/
Unmerged revisions
- 154. By Skyler Berg on 2015-08-20
-
Add nfs-mount-options options configuration option

charm_lint_check #8427 nova-compute-next for sberg-l mp268680
LINT OK: passed
Build: http:// 10.245. 162.77: 8080/job/ charm_lint_ check/8427/