See this blog post for a detailed walkthrough together with a video demonstrating usage.
Amun is a service that executes the given application stack in the requested environment - given the list of package that should be installed as well as given the hardware that is requested to run the application. Its primary purpose is to act as an execution engine for Thoth where applications are built and tested (applications are automatically generated given the software requirements). However, it can be used to verify and check application behavior in a cluster.
There are performed 2 core steps by Amun:
- Assemble/build the given software by installing requested native packages and/or Python packages into a container image. OpenShift's ImageStreams and builds are used under the hood.
- Execute the given application using a script that is provided by a user - there are run "inspection jobs" that execute provided user script. All the information related to the node where the inspection job was run are aggregated. Such information consists of hardware available (such as CPU, CPU flags and features, and such; see amun-hwinfo) as well as information from the kernel's process control block (such as number of context switches performed, time spent in user/kernel space and such).
The second step is performed if the build succeeded and a user provided a script to test the application with in the given build environment on the requested hardware (there are used node selectors in the cluster for this purpose).
The actual second step is used to gather information whether the application runs with packages being installed in the build step as well as information such as performance characteristics or any other runtime-related information of the assembled application.
As Amun accepts purely JSON on its input, the inspection step requires a test file that is written to disk with an execute flag and run.
All the relevant logs from build and inspection job runs are aggregated and stored on Ceph together with actual results of inspections.
See thoth-station/performance repository for an example of a script that can be executed on Amun.
One can see Amun as a CI running in a cluster.
- thoth-station/performance
- Dependency Monkey ZOO
- thoth-station/datasets
- thoth-station/performance
- Kaggle dataset
A single request to API is composed of:
- an identifer of the inspection
- a base image itself (e.g. ubi8)
- a list of native packages (RPM or Deb packages) that should be installed into the requested base image
- a list of Python packages that should be installed into the requested base
image in a form of
Pipfile/Pipfile.lock - a script (bash, Python or any other scripting language - if the given environment knows how to execute the script; if it has required interpreter)
- hardware requirements for pod placement performing builds of application stack (installing necessary dependencies)
- hardware requirements for pod placement performing actual application execution - "inspection jobs"
See provided OpenAPI/Swagger specification available in this Git repository. base image is required parameter.
Upon a successful request to Amun API, a user obtains an inspection_id.
This identifier is used to reference the given request. On the build endpoints
there are leveraged information about build status and the actual build logs,
on the job endpoints, there are leveraged information about the actual
inspection runs - logs and logs printed to standard output and standard error
stream. These results are obtained on a successful inspection run.
Each time there is created a request with a script run (so there is actually spawned job responsible for running the provided script), there is run an init container that gathers information about hardware that is present on node where the application is run. This information is available in a form of JSON and becomes part of the actual result of an inspection run.
The Python script that gathers information about hardware present can be found in amun-hwinfo repository.
I, as an Amun user, would like to test performance of optimized TensorFlow builds available on the AICoE Python package index. I would like to use:
- TensorFlow provided on AICoE index (provide a
PipfileandPipfile.lockrespecting Pipenv configuration to use different package indexes) - Python3, CUDA in specific version, .. - installed as RPMs
- use a cluster node that exposes a GPU with CUDA support
- I would like to use ubi8 as a base image
- I don't need a node with GPU support to assemble/build the TensorFlow application
- I provide a Python script that is a TensorFlow application run to gather information about TensorFlow (the application can print a JSON with results, but can also push data to a remote API stated in the Python script itself).
Amun guarantees the application is built as requested and it is placed on the correct node inside the cluster given the application requirements (a GPU with CUDA support).
If the build part fails, the script cannot be run. The build failures can be observed on exposed build endpoints and are available on Ceph.
To comunicate with Amun API, use the autogenerated Swagger client that available in the amun-client repository.
Library called thoth-storages implements adapters that offer Python interface for accessing inspection files.
An example of an Amun inspection request can be found in this repo in examples directory. The structure corresponds to one inspection placed on Ceph with all the relevant bits:
── inspection-rhtf-conv2d-0f845f38 # inspection id
├── build
│ ├── Dockerfile
│ ├── log
│ └── specification
└── results
├── 0
│ ├── hwinfo
│ ├── log
│ └── result
├── 1
│ ├── hwinfo
│ ├── log
│ └── result
└── 2
├── hwinfo
├── log
└── result<inspection-id>/build/Dockerfile- automatically generated Dockerfile used to build the application, this Dockerfile is a transcription of the JSON input to Amun API<inspection-id>/build/log- build log produced during the application assembling<inspection-id>/build/specification- whole input (toghether with defaults) sent to Amun API endpoints that forms Amun inspection request
All results are indexed, index corresponds to one of the items in the inspection batch (multiple inspection job runs can be performed with each request so the application is built just once).
<inspection-id>/results/<num>/hwinfo- information about hardware on which the inspection job was run (see thoth-station/amun-hwinfo)<inspection-id>/results/<num>/log- inspection run log -- standard error and standard output as produced by the inspection script (or any library/subprocess it uses).<inspection-id>/results/<num>/result- the actual result of an inspection run together with process information from the kernel.
All the results are available on Ceph or any object storage providing AWS S3 compatible interface.
It's possible to observe how inspections proceed using Argo UI. Argo UI is exposed on deployment.
All manifests required to deploy this application are available in thoth-station/thoth-application.
See this video with instructions on how to deploy Amun into a cluster using kustomize.
This software is released under the terms of GNU General Public License in version 3.
© Red Hat; AICoE team - Project Thoth