This commit is contained in:
Fred Park 2019-12-13 21:32:26 +00:00
Родитель 85985d6a3e
Коммит d6da749f9c
Не найден ключ, соответствующий данной подписи
Идентификатор ключа GPG: 3C4D545F457737EB
3 изменённых файлов: 31 добавлений и 3 удалений

Просмотреть файл

@ -2,8 +2,27 @@
## [Unreleased] ## [Unreleased]
## [3.9.1] - 2019-12-13
### Added
- Support `--no-wait` on pool creation to allow the command to skip
waiting for the pool to become idle
- Allow ability to ignore GPU warnings, please see pool configuration
- Ubuntu 18.04 SR-IOV IB/RDMA Packer script
### Changed
- **Breaking Change:** improved per-job autoscratch setup. As part of this
change the `auto_scratch` property in the jobs configuration has changed.
- Provide ability to setup via dependency or blocking behavior
- Allow specifying the number of VMs to span
- Allow specifying the per-job autoscratch task id
- Allow multiple multi-instance tasks per job in non-`native` mode
- Update GlusterFS version to 7
### Fixed ### Fixed
- Fix merge task regression with enhanced autogenerated task id support - Fix merge task regression with enhanced autogenerated task id support
- Fix job schedule submission regression
([#329](https://github.com/Azure/batch-shipyard/issues/329))
- Fix per-job autoscratch provisioning due to upstream dependency changes
## [3.9.0] - 2019-11-15 (SC19 Edition) ## [3.9.0] - 2019-11-15 (SC19 Edition)
### Added ### Added
@ -1707,7 +1726,8 @@ transfer is disabled
#### Added #### Added
- Initial release - Initial release
[Unreleased]: https://github.com/Azure/batch-shipyard/compare/3.9.0...HEAD [Unreleased]: https://github.com/Azure/batch-shipyard/compare/3.9.1...HEAD
[3.9.1]: https://github.com/Azure/batch-shipyard/compare/3.9.0...3.9.1
[3.9.0]: https://github.com/Azure/batch-shipyard/compare/3.8.2...3.9.0 [3.9.0]: https://github.com/Azure/batch-shipyard/compare/3.8.2...3.9.0
[3.8.2]: https://github.com/Azure/batch-shipyard/compare/3.8.1...3.8.2 [3.8.2]: https://github.com/Azure/batch-shipyard/compare/3.8.1...3.8.2
[3.8.1]: https://github.com/Azure/batch-shipyard/compare/3.8.0...3.8.1 [3.8.1]: https://github.com/Azure/batch-shipyard/compare/3.8.0...3.8.1

Просмотреть файл

@ -22,4 +22,4 @@
# FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER # FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
# DEALINGS IN THE SOFTWARE. # DEALINGS IN THE SOFTWARE.
__version__ = '3.9.0' __version__ = '3.9.1'

Просмотреть файл

@ -28,7 +28,7 @@ It is **strongly recommended** to use Shared Image Gallery resources instead
of directly using an Azure Managed Image for increased reliability, robustness of directly using an Azure Managed Image for increased reliability, robustness
and performance of scale out (i.e., pool allocation with target node counts and performance of scale out (i.e., pool allocation with target node counts
and resize up) operations with Azure Batch pools. These improvements hold even and resize up) operations with Azure Batch pools. These improvements hold even
for Shared Image Gallery resource with a replica count of 1. for a Shared Image Gallery resource with a replica count of 1.
This guide will focus on creating Shared Image Gallery resources for use with This guide will focus on creating Shared Image Gallery resources for use with
Azure Batch and Batch Shipyard. Azure Batch and Batch Shipyard.
@ -263,12 +263,20 @@ and the required user-land software for Infiniband installed. It is best to
base a custom image off of the existing Azure platform images that support base a custom image off of the existing Azure platform images that support
Infiniband/RDMA. Infiniband/RDMA.
#### MPI Libraries
If you are utilizing MPI, the associated runtime(s) must be installed such
that they are invocable by the calling programs.
#### Storage Cluster Auto-Linking and Mounting #### Storage Cluster Auto-Linking and Mounting
If mounting a storage cluster, the required NFSv4 or GlusterFS client tooling If mounting a storage cluster, the required NFSv4 or GlusterFS client tooling
must be installed and invocable such that the auto-link mount functionality must be installed and invocable such that the auto-link mount functionality
is operable. Both clients need not be installed unless you are mounting is operable. Both clients need not be installed unless you are mounting
both types of storage clusters. both types of storage clusters.
#### Per-Job Autoscratch
If utilizing the per-job autoscratch feature, then BeeGFS Beeond must be
installed so that a shared file system can be created.
#### GlusterFS On Compute #### GlusterFS On Compute
If a GlusterFS on compute shared data volume is required, then GlusterFS If a GlusterFS on compute shared data volume is required, then GlusterFS
server and client tooling must be installed and invocable so the shared server and client tooling must be installed and invocable so the shared