Live migration is currently totally broken if a NUMA topology is present. This affects everything that's been regrettably stuffed in with NUMA topology including CPU pinning, hugepage support and emulator thread support. Side effects can range from simple unexpected performance hits (due to instances running on the same cores) to complete failures (due to instance cores or huge pages being mapped to CPUs/NUMA nodes that don't exist on the destination host). Until such a time as we resolve these issues, we should alert users to the fact that such issues exist. A workaround option is provided for operators that _really_ need the broken behavior, but it's defaulted to False to highlight the brokenness of this feature to unsuspecting operators. Change-Id: I217fba9138132b107e9d62895d699d238392e761 Signed-off-by: Stephen Finucane <sfinucan@redhat.com> Related-bug: #1289064
6.1 KiB
Feature Classification
This document presents a matrix that describes which features are ready to be used and which features are works in progress. It includes links to relevant documentation and functional tests.
Warning
Please note: this is a work in progress!
Aims
Users want reliable, long-term solutions for their use cases. The feature classification matrix identifies which features are complete and ready to use, and which should be used with caution.
The matrix also benefits developers by providing a list of features that require further work to be considered complete.
Below is a matrix for a selection of important verticals:
matrix-gp
matrix-nfv
matrix-hpc
For more details on the concepts in each matrix, please see notes-on-concepts
.
General Purpose Cloud Features
This is a summary of the key features dev/test clouds, and other similar general purpose clouds need, and it describes their current state.
Below there are sections on NFV and HPC specific features. These look at specific features and scenarios that are important to those more specific sets of use cases.
NFV Cloud Features
Network Function Virtualization (NFV) is about virtualizing network node functions into building blocks that may connect, or chain together to create a particular service. It is common for this workloads needing bare metal like performance, i.e. low latency and close to line speed performance.
HPC Cloud Features
High Performance Compute (HPC) cloud have some specific needs that are covered in this set of features.
Notes on Concepts
This document uses the following terminology.
Users
These are the users we talk about in this document:
- application deployer
-
creates and deletes servers, directly or indirectly using an API
- application developer
-
creates images and apps that run on the cloud
- cloud operator
-
administers the cloud
- self service administrator
-
runs and uses the cloud
Note
This is not an exhaustive list of personas, but rather an indicative set of users.
Feature Group
To reduce the size of the matrix, we organize the features into groups. Each group maps to a set of user stories that can be validated by a set of scenarios and tests. Typically, this means a set of tempest tests.
This list focuses on API concepts like attach and detach volumes, rather than deployment specific concepts like attach an iSCSI volume to a KVM based VM.
Deployment
A deployment maps to a specific test environment. We provide a full description of the environment, so it is possible to reproduce the reported test results for each of the Feature Groups.
This description includes all aspects of the deployment, for example the hypervisor, number of nova-compute services, storage, network driver, and types of images being tested.
Feature Group Maturity
The Feature Group Maturity rating is specific to the API concepts, rather than specific to a particular deployment. That detail is covered in the deployment rating for each feature group.
Note
Although having some similarities, this list is not directly related to the DefCore effort.
Feature Group ratings:
- Incomplete
-
Incomplete features are those that do not have enough functionality to satisfy real world use cases.
- Experimental
-
Experimental features should be used with extreme caution. They are likely to have little or no upstream testing, and are therefore likely to contain bugs.
- Complete
-
For a feature to be considered complete, it must have:
- complete API docs (concept and REST call definition)
- complete Administrator docs
- tempest tests that define if the feature works correctly
- sufficient functionality and reliability to be useful in real world scenarios
- a reasonable expectation that the feature will be supported long-term
- Complete and Required
-
There are various reasons why a complete feature may be required, but generally it is when all drivers support that feature. New drivers need to prove they support all required features before they are allowed in upstream Nova.
Required features are those that any new technology must support before being allowed into tree. The larger the list, the more features are available on all Nova based clouds.
- Deprecated
-
Deprecated features are those that are scheduled to be removed in a future major release of Nova. If a feature is marked as complete, it should never be deprecated.
If a feature is incomplete or experimental for several releases, it runs the risk of being deprecated and later removed from the code base.
Deployment Rating for a Feature Group
The deployment rating refers to the state of the tests for each Feature Group on a particular deployment.
Deployment ratings:
- Unknown
-
No data is available.
- Not Implemented
-
No tests exist.
- Implemented
-
Self declared that the tempest tests pass.
- Regularly Tested
-
Tested by third party CI.
- Checked
-
Tested as part of the check or gate queue.
The eventual goal is to automate this list from a third party CI reporting system, but currently we document manual inspections in an ini file. Ideally, we will review the list at every milestone.