From Fedora Project Wiki

Revision as of 17:50, 28 February 2014 by Mattdm (talk | contribs) (update wip from etherpad)


Working on this now at http://cloudpad-samkottler.rhcloud.com/p/CloudChanges

Suggested Structure

Basically, each of these are going to become a Change using the Fedora Features process (Changes/Policy). So, each should be a very lightweight version of Changes/EmptyTemplate. We should also rate each one for importance (overall/strategically) and urgency (things that need to get done *now*, things that need to land in f21, things that can land after) + a rationale for the urgency (blocks something, competitive need, etc). So, I'm envisioning a list here which looks like this:


Change: Change Name

Summary: Elevator pitch for the change. Importance: [ vital | moderate | nice-to-have ] Timeframe: [ now | F21 alpha | F21 beta | F21 release | F22 | F23+ ] / reason (e.g. "blocks ability to produce images" ) Scope: [ self-contained | complex system-wide ] link to change proposal page (can be not actually filled out yet)


Additionally, we are going to have dependencies on changes that are largely in other Fedora groups. These aren't really our changes, so we shouldn't be submitting them, but we should also list them. The enhanced automatic QA features of taskotron would be an example. Web design and marketing are others. So, something like:


External Need: Dependency Name

Summary: Elevator pitch for the change. Importance: [ vital | moderate | nice-to-have ] Timeframe: [ now | F21 alpha | F21 beta | F21 release | F22 | F23+ ] / reason (e.g. "blocks ability to produce images" ) Fedora Sub-Project/SIG: [ QA | Websites | etc ] Cloud SIG owner: (We can't just demand things, obviously. Who from our group is going to interface and possibly contribute?) link to relevant whatever


REAL STUFF STARTS BELOW HERE

The List


Change: Move to ImageFactory For image Creation

Summary: Create images using Anaconda in Koji rather than obsolete, unmaintained appliance-creator. Allows fedmsg integration for upload service, and also could produce official Docker images. Importance: vital (basic image creation is nice-to-have, fedmsg integration is important to automation, and producing Docker images is strategically vital) Timeframe: at least a month lead time before F21 alpha / If we are going to release images made this way, we need them made that way for early testing, and there will be some porting work to the kickstart Scope: self-contained (work with Rel Eng) https://fedoraproject.org/wiki/Changes/ImageFactory_in_Koji


External Need: Automatic Image Upload

Summary: Whenever an image is built, it is automatically uploaded to our cloud provider targets (EC2, etc.) and to the fedora ftp site (and possibly mirrors if we can convince mirror admins to drink from that firehose) Importance: vital (current process where rel eng does it by hand will not scale) Timeframe: whenever we can get it / this adds value whenever it happens Fedora Sub-Project/SIG: Release Engineering, possibly Infrastructure for resources Cloud SIG owner: ???????? (note summer intern may help with that, if it is not too late) (need link to something further -- possibly we file this as a change)


External Need: Automatic Smoketests on Image Build

Summary: When a new image is built in Koji, automatically boot it and run a basic matrix of tests. Importance: moderate (worst case, we can keep doing this by hand) Timeframe: F21 alpha / Want to reduce manual test workload Fedora Sub-Project/SIG: QA and the Taskotran project Cloud SIG owner: ???????? https://fedoraproject.org/wiki/User:Tflink/taskotron_development_plan


External Need: Scratch Builds on Change

Summary: Whenever the kickstart changes, _or_ an RPM which is on the image hits the tree, a new scratch image is automatically built. Importance: nice to have (makes development much easier, and makes it quick to spot and fix problems before they affect anyone) Timeframe: whenever we can get it / this adds value whenever it happens Fedora Sub-Project/SIG: Release Engineering, possibly Infrastructure for resources Cloud SIG owner: ???????? (need link to something further -- possibly we file this as a change)


External Need: Updated Web Site

Summary: Since we are now one of the three top level artifacts Fedora produces, we want to emphasize our unique niche. Updated website with new flashier branding, plus tools for selecting different images for different use cases Importance: vital (it's basically part of the whole exercise Timeframe: F21 release / arguably, having the web site up _is_ the release. And it'd be nice to have earlier, like at the alpha and beta Fedora Sub-Project/SIG: Websites Cloud SIG owner: Joe Brockmeier https://fedorahosted.org/fedora-websites/ticket/248


Change: Docker Host Image

Summary: Fedora Cloud agreed to make a base image plus several tailored to specific purposes. This is one of the tailored ones — Docker host ready to go. Importance: moderate (Docker is hot stuff and we need an answer to CoreOS, but if we fail to do it, we are not any worse-off than we are now, and Docker can still be used on the base iamge.) Timeframe: F21 alpha / If it's not ready for alpha, we have missed this release Scope: self-contained https://fedoraproject.org/wiki/Changes/Docker_Cloud_Image


Change: Big Data Image

Summary: Fedora Cloud agreed to make a base image plus several tailored to specific purposes. This is one of the tailored ones, produced in collaboration with the Big Data SIG. Importance: nice to have (if we fail to do it, we are not any worse-off than we are now, and big data tools can still be used on the base iamge.) Timeframe: F21 alpha / If it's not ready for alpha, we have missed this release Scope: self-contained https://fedoraproject.org/wiki/Changes/Big_Data_Cloud_Image


Change: OpenShift Out-of-the-Box Image

Summary: Fedora Cloud agreed to make a base image plus several tailored to specific purposes. This is one of the tailored ones — OpenShift ready to run. Importance: nice to have (if we fail to do it, we are not any worse-off than we are now, and OpenShift can still be installed on the base iamge.) Timeframe: F21 alpha / If it's not ready for alpha, we have missed this release Scope: self-contained https://fedoraproject.org/wiki/Changes/Big_Data_Cloud_Image


Change: Docker Container Image

Summary: This is Fedora running inside Docker. Currently, there are non-official images in the docker index, but we'd like them to really come out of release engineering. Importance: nice to have (The unofficial image situation is not ideal. Plus, this is an area where we can show leadership.) Timeframe: F21 alpha / If it's not ready for alpha, we have missed this release Scope: self-contained https://fedoraproject.org/wiki/Changes/Docker_Container_Image

Change: (A)Periodic Updates to the Images

Summary: We want to be able to release updated images not just at release time. Hope for a one-month regular cadence, plus emergency updates if needed. Importance: vital (People don't like to yum update their images, and since we change so much, doing so is a significant launceh delay. let's make it so they don't have to, but still get updates) Timeframe: F21 final + 1 month / Obviously better if we get things lined up earlier Scope: self-contained https://fedoraproject.org/wiki/Changes/(A)Periodic_Updates_to_Images


External Need: Batched Updates

Summary: We want to produce updated images on a monthly cadence. It would be nice if we could produce those from QA'd bunches of packages. Importance: moderate (it will be hard to implement image refresh without this, but we could do it) Timeframe: F21 release + 1 month / Obviously better if we get things lined up earlier Fedora Sub-Project/SIG: Primarily QA, but Rel Eng and Infrastructure too. This is pretty big. Cloud SIG owner: ???????? (this probably needs someone actively contributing to initial and ongoing work) http://flock2013.sched.org/event/8c4f702e42814598e0e4e31b188a0ae6


Change: Install without i10n / l18n support (w/optional install)

Summary. We need to have a cloud image without the overhead/extra space requirements of i10n / l18n support, but want to provide the option to install these packages if needed. In many cases these are not necessary for images running in the cloud. Importance. moderate (Should not be overly difficult to implement, but is not a show-stopper if it's not complete by F21.) Timeframe. F21 alpha / should be at least mostly implemented by alpha or won't be ready for this release. Scope. self-contained Cloud SIG owner. mattdm (I already have a conversation open with anaconda and the packaging team) https://fedoraproject.org/wiki/Changes/Optional_i10n_and_l18n_support_cloud_image


Change: Install without documentation (w/optional install later)

Summary. In many cases, users will not want documentation on cloud images. We should be able to provide support to create/deliver cloud images without documentation but still have it available for installation if desired. Importance. moderate Timeframe. F21 alpha / should be at least mostly implemented by alpha or will not be ready for this release. Scope. self-contained Cloud SIG owner. ??????? https://fedoraproject.org/wiki/Changes/Optional_documentation_cloud_image


External Need: Software Collections for Cloud Users

Summary: Importance: vital (provides a meaningful reason to use Fedora cloud image, and helps insulate against rapid change) Timeframe: F21 release / This is a requirement of users in production. Fedora Sub-Project/SIG: Environments and Stacks Cloud SIG owner: ???????? link to relevant whatever


Change: Cloud-Friendly Kernel Packaging'

Summary: Modules that are not necessary in virtualized environments become optionally installable Importance: nice-to-have Timeframe: F21 beta / need widespread testing to make sure nothing really needed is left out Scope: self-contained (kernel team with input/feedback from cloud) https://fedoraproject.org/wiki/Changes/Modular_Kernel_Packaging_for_Cloud


Change: Modularized SELinux Packaging'

Summary: Monolith SELinux policy has rules for everything that could exist, not just that which is installed. Possible room for reduction here. Importance: nice-to-have Timeframe: F21 or F22 / This is not low-hanging fruit. Scope: self-contained (SELinux) (No Feature filed at this time -- discuss with SELinux policy maintainers)