A application for triggering new builds and jobs for tensorflow-build releases.
Build status of continus integration i.e. Quay. -
$ oc create --filename openshift/job_template.yaml
$ oc new-app --template=tensorflow-release-job -p OCP_SECRET=<secret> -p BUILD_MAP="$(cat config.json)" -p SESHETA_GITHUB_ACCESS_TOKEN=<GITHUB_TOKEN>
- The build map config.json contains the python version and os version for which tensorflow-build is to be triggered.
-
Structure of config.json:
-
OS Version specific variables can be specified inside the os_version value with Upper Case Key variable(required).
Ex: "TF_NEED_HDFS": "{1|0}"{ "python_version": { "os_version": { "S2I_IMAGE": "os_registry", "BAZEL_VERSION": "bazel_version", "TF_NEED_CUDA": "{1|0}", "RESOURCE_LIMITS_CPU": "4", "RESOURCE_LIMITS_MEMORY": "8Gi" } } }
-
-
Check for the Resource Quota:
-
The functionality of resource quota check while triggering builds and jobs is provided in the application. Required variables to set:
To Set the name of the resource quota:
QUOTA_NAME = <resource_quota_name> (default: <namespace>-quota )
To Disable the check for resource quota:
RESOURCE_QUOTA = 0 (default: 1)
(Pass it as a parameter in Step 2 of Deployment)
-
-
The OCP_SECRETS are the openshift variables:
- OCP_URL =
<openshift_url | ex: https://paas.upshift.redhat.com>
- OCP_NAMESPACE =
<openshift_namespace | ex: thoth-station>
- OCP_TOKEN =
<openshift_token>
(Use Service account token for production | For Testing , Session Token can be used(As these have 24hr life))
Store the above information in secret and pass it to the appliction as parameter(shown in step-2 of Deployment).
- OCP_URL =
-
Create SECRET in openshift:
$ oc create secret --namespace "{{ OCP_NAMESPACE }}" generic {{OCP_SECRET}} \ --from-literal=OCP_URL="{{ OCP_URL }}" \ --from-literal=OCP_TOKEN="{{ OCP_TOKEN }}" \ --from-literal=OCP_NAMESPACE="{{ OCP_NAMESPACE }}" \ --type=opaque
- All the tensorflow build related parameters can be passed to step-2 of Deployment as parameters.
(Default: fedora28 python36)
- All the tensorflow build related parameters can be passed to step-2 of Deployment as parameters.