-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/45439/
-----------------------------------------------------------

(Updated March 31, 2016, 4:48 p.m.)


Review request for mesos, Ben Mahler, Rob Todd, and Vikrama Ditya.


Changes
-------

Updated based on offline feedback from bmahler.

The biggest change was to manage two lists of GPUs instead of one (i.e., we now 
use one list for the total-available GPUs in the system and another 
per-container list of the GPUs that container currently owns).

We also removed the helpers for `prepareCgroupsDevicesWhitelist` and 
`allow/denyGPpu` in favor of inlining their functionality.

We also did a `s/ListEntry/Entry/` to update based on the `cgroups::devices` 
commits that landed.

Other than that the rest of the changes were mostly around comments and 
`Error/Failure` string formatting.


Summary (updated)
-----------------

Completed MVP implementation of the Nvidia GPU isolator.


Bugs: MESOS-4625
    https://issues.apache.org/jira/browse/MESOS-4625


Repository: mesos


Description
-------

The current implementation includes some code for generic device
isolation. Once we start isolating more devices, we should abstract
this out and figure out an architecture to support "sub-device"
isolation.

This was joint work with Rob Todd from Nvidia <rt...@nvidia.com>.


Diffs (updated)
-----

  src/slave/containerizer/mesos/isolators/cgroups/devices/gpus/nvidia.hpp 
1e17df1f7d1daf913e0c03e0290dd85b51c2ade0 
  src/slave/containerizer/mesos/isolators/cgroups/devices/gpus/nvidia.cpp 
c2cdc8fde7a85741be6494ea664d3719d1f13a43 

Diff: https://reviews.apache.org/r/45439/diff/


Testing
-------

Test to come in subsequent patch.


Thanks,

Kevin Klues

Reply via email to