Move from phases to jobs and use bring your own cloud pool #3908
Conversation
|
Just cherry-pick you changes to my branch and trigger a new build. Let's see if we can publish packages then. I will discard your changes on my side later so that we can have your changes from the right source. Thanks! |
There was a problem hiding this comment.
Thanks now I can build ML.NET with the latest OnnxRuntime nuget. Note that my tests fails due to some random effects, so I trigger the two builds again.
[Update] Hold on. The build didn't really produce Nuget packages because two tests failed.
[Update] We can now build under Release mode for .NET Core 3.0. Debug mode build still fails because we need to increate build time. Consequently, I think this PR can be checked in now.
[Update] After increasing the build time limit to 40 mins, both .NET Core 3.0 builds passes.
|
@codemzs @wschin - I think moving from phases to jobs should happen in the official build as well: https://github.com/dotnet/machinelearning/blob/master/build/vsts-ci.yml. I've opened #3926 for this. |
* Move from phases to jobs and use bring your own cloud pool * Fix queue * Fix build
ghost
locked as resolved and limited conversation to collaborators
This is needed to unblock: #3881
Basically the hosted azure devops machines have limited hard drive (25gb) and since tests are executed on the build machine it is running out of space. BYOC pools have 512 gb so that will unblock this scenario.
However I do think, machinelearning repo should move to run tests on helix in the long term.
Note: this is easier to review if omitting whitespaces from the diff
cc: @codemzs @wschin