[RELENG-59] Update CSIT: Remove featuresBoot and install from karaf shell Created: 07/Jun/17 Updated: 28/Feb/18 |
|
| Status: | Open |
| Project: | releng |
| Component/s: | Jenkins Job Builder |
| Affects Version/s: | unspecified |
| Fix Version/s: | None |
| Type: | Bug | ||
| Reporter: | A H | Assignee: | Unassigned |
| Resolution: | Unresolved | Votes: | 0 |
| Labels: | None | ||
| Remaining Estimate: | Not Specified | ||
| Time Spent: | Not Specified | ||
| Original Estimate: | Not Specified | ||
| Environment: |
Operating System: All |
||
| External issue ID: | 8627 |
| Description |
|
Change the way we install features for CSIT. We are not going to use featuresBoot any longer and install from the karaf shell: https://lists.opendaylight.org/pipermail/release/2017-June/011157.html |
| Comments |
| Comment by Daniel Farrell [ 08/Jun/17 ] |
|
Configuration management tools for ODL, like ansible-opendaylight and puppet-opendaylight, currently use the featuresBoot line to configure which features are installed. This logic is consumed by major users, like OPNFV. It's critical that we smoothly transition that config logic to use whatever the new method is. I have a Int/Pack Trello tacking this here: |
| Comment by Vratko Polak [ 08/Jun/17 ] |
|
During Developer Design Forum we were discussing this, and we have decided that the installation scripts shall stay in Releng/Builder for Nitrogen cycle. |
| Comment by Jamo Luhrsen [ 08/Jun/17 ] |
|
one main reason this option was discussed was because if features are there |
| Comment by Daniel Farrell [ 09/Jun/17 ] |
|
My main concern about removing featuresBoot is that config management tools are much better at working with files than interactive sessions. They can check the status of a file and see if it needs to be updated, or see if it has changed and do some action. They can also work with the file before ODL is started (after an RPM install, before `sudo systemctl start opendaylight`). It's not clear how well we can do this with only interactive commands. After OPNFV Summit I can try to put together a PoC of how we could migrate config mgmt tools to be interactive, see how horrible it actually turns out to be. My other concern is that this seems like a bug in Karaf. Both install methods should use the same feature resolver, or generally if both are provided, it should be possible to mix them. |
| Comment by Luis Gomez [ 09/Jun/17 ] |
|
I have to agree with Daniel here, if both install methods are not compatible anyomore it sounds like a karaf regression, we should at least reach the upstream community to understand better why this change if it is not a bug. |
| Comment by Vratko Polak [ 29/Jun/17 ] |
|
So far it seem like featuresBoot keeps working in Karaf 4, at least when there are no feature:install commands executed at runtime. There may be some CSIT which need this improvement, but most of CSIT jobs are fine with the current releng/builder scripts. |