-
Notifications
You must be signed in to change notification settings - Fork 391
Add a streamlined deploy option through ansible #155
Conversation
… expected to create the deploy/files/topology.json file and create an inventory file for ansible. The deploy playbook will upgrade OS packages, install glusterfs-client, build heketi, run gk-deploy, and create the StorageClass.
…perty to gk-deploy.
…le. Ansible now has a GK_DIR variable that points to the project root deploy directory.
…onsistency. Removed unnecessary ansible groups in inventory.ini.sample. Added the inventory and topology to .gitignore.
…hould just copy both template directories and let gk-deploy do the auto-detection. Renamed destroy_vg.yaml to destroy_vgs.yaml.
…d on. Updated code references to use .yaml files instead of .yml.
…o get the heketi service IP and port.
I want to do a careful review of this and test it out myself, so it may take me a bit. But I like it thus far! |
@ravishivt A bit of a non-update, since I don't want you to think I'm not thinking about this: I've been prepping for a cross-continental trip to two conferences, so my apologies for the continued delay. I'll have plenty of time on planes and a week between the conferences to try and check this out, so I hope to get some comments back to you soon. |
@jarrpa No worries. Take your time and safe travels. One thing to note is that your ansible version must be < 2.2.x. Ansible 2.x and after has a bug that our playbook will hit. |
@ravishivt Heyo. Coming in almost a month later to say that i still haven't been able to take this for a proper test run. :( My apologies. I've been hit with deadlines downstream, but we're hoping to clear them by next week. I have, however, been able to at least read through most of it and offhand it looks pretty solid on the Debian side. I'm currently not planning to run the Debian side through its paces, mostly because I'd rather save the time of figuring out how to bring up an Ubuntu cluster (it's been long enough as it is!). One thing to note is that with the release of heketi version 4 I plan on going off of that version for this project rather than tracking heketi master. That work will go in to the repo before this PR does, which means I'd like this playbook to build heketi v4 rather than master. I'll update this PR when that happens, and if I'm a decent human being maybe even provide patches. :) Again, my apologies, we really do appreciate your contributions and don't like how long we've been letting this sit idle. Hope to bring some good news soon. |
@ravishivt All right, I've taken a crack at this and have some questions/comments:
|
I wonder if you could provide a heketi-cli container as I cant build it on an Fedora Atomic Host |
@goern how does that request/question relate to this PR? It would certainly be possible. up to now we've never needed to build a specific cli container. |
Closing this PR. It can be reopened if needed. |
This implements #149. It creates a new ansible/ directory with ansible playbooks and roles to deploy the project to a pre-existing cluster. The user needs to still specify the ansible inventory file and the topology.json file. The new ansible option does not supersede the existing manual deployment procedure or documentation. An initial README is added with details on what the playbooks and roles do and how to use them.
This change is![Reviewable](https://camo.githubusercontent.com/23b05f5fb48215c989e92cc44cf6512512d083132bd3daf689867c8d9d386888/68747470733a2f2f72657669657761626c652e696f2f7265766965775f627574746f6e2e737667)