Giters

riebling / acai

Notebook for ramping up on ACAI project

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

acai

Notebook for ramping up on ACAI project

Tasks

  1. Meet with Eric Nyberg for Overview DONE
  2. Meet Cameron Dashti for intro to The Project Zone DONE
  3. Current task: read through TA materials on TPZ wiki : 40% complete DONE
  4. Meet Eric & Boyue to discuss running his ACAI container in TPZ DONE
  5. lots of trying and lots of failures running Boyue's code in an Amazon VM (see http://github.com/liboyue/BOOM issues)
  6. Cameron supplies template for submitter (grader to come) DONE

terminology

ENA, ENI - Elastic Network Adapter, Elastic Network Interface  
HVM - Hardware Virtual Machine
AGS - auto grader system (?)
EMR - elastic map reduce
Kubernetes - Tool to scale, deploy lots of containers
Terraform - Provision dependencies in cloud (like Vagrant/Chef/Puppet)
primer = ?
MOSS - "measure of software similarity" cheat detection for code

Submitting Code to Autograder

File structure needs to be observed for certain required elements;
These go into a tarball and are uploaded to TPZ Data goes into well-known Amazon S3 buckets e.g. S3://cmucc-code/s18/15719/spark-etl/

AssessMes (quizzes) HOWTO

Done using TPZ Assessment Portal
Sets of 3 subsets of 3 similar questions (to discourage brute forcing)

TPJ - The Project Zone

  • Logging into TPZ uses CMU/Andrew authentication rather than it's own username+password, but because it used to, signing in can be problematic, for example when creating AssessMe quizzes

A TPZ "Writeup" - instructional materials

  • Lives in folder e.g: Project/writeup-xxx/010_writeup_name.md

TPZ Submissions

  • Require a password, generated on TPZ site (page?)

AWS stuff

  • The linux API to AWS EC2 is a command 'aws'. It's essential, and useful, but overloaded. There are 155 subcommands, or "things you can do" with ami
  • Master AWS account (root user) uses IAM (Identity Access Manager) roles to grant sub-account permissions
  • the ways you log into AWS differ for master vs. IAM accounts, root user cannot switch roles

Random Notes/Questions

Q: How to know what's already running in the cloud to submit for autograder?

N: cloud computing course uses us-east-1

What All is Needed to Submit to Autograder

Process happens in 2 steps: submitter, checker

  • TPZ accounts (faculty / staff / students)

    • TPZ_USERNAME
    • TPZ_PASSWORD

  • Andrew ID

  • Course ID within TPZ, e.g. we have F18: Design & Engineering of Intelligent Information Systems
    F18 11-791 Design & Engineering of Intelligent Information Systems

    • Projects within TPZ
      • Module (homework) within TPZ
        • Open time
        • Deadline
        • Writeup
        • (optional) AssessMe
        • generated TPZ Submission Password

  • AWS accounts (faculty / staff / students)

    • IAM Users
    • Policies for users, e.g TPS staff needs to set us up withAmazonEC2ContainerRegistryReadOnly policy
    • a profile with role set up AWS_PROFILE=assumed_role so as to be able to run
    eval $(aws ecr get-login --no-include-email --region us-east-1)
docker pull 939223853384.dkr.ecr.us-east-1.amazonaws.com/cmucc/ags:ubuntu
docker tag 939223853384.dkr.ecr.us-east-1.amazonaws.com/cmucc/ags:ubuntu cmucc/ags:ubuntu

  • a submitter/ folder (where does this need to be, where does it run?)

    • Makefile
    • SHC_DOCKER_IMAGE=cumcc/shc (what/where is this exactly?)
    • SUBMITTER_NAME=submitter and corresponding submitter/submitter.sh
    • CHECKER_NAME=checker and corresponding submitter/checker.sh

  • a grader/ folder (when, where, and by whom does this need to be / run?)

    • Makefile
    • ags-image-pull.sh
    • pom.xml

  • we have to write our own submitter and checker code (in Java?)

    • a local copy of the AGS Docker image
    • Submitter Submitter HowTo mentions "Compile the Submitter with Docker and Makefile", which is very confusing. It talks about cloning the shc git repo, building it, running it to turn submitter.sh (do we have to write this?) into submitter.sh.x.c - it shows some example code that does not seem to have anything to do with Docker or Makefile. This seems like notes or "extra things to remember" regarding a Submitter, but not at all a tutorial or step-by-step example on how to actually create and use Submitters.

  • Grader grader HOWTO

    • seems like developing a grader is it's own project, with learning curve, iterative development, testing, before deplying:
    • There are so many prerequisites just to develop a grader, including use cases of
    • download, set up, and run AGS image locally as a Docker container
    • develop Java grader
    • run interactively
    • evaluate results:
      • feedback
      • log
      • score
    • practice submitting
    • validate submissions
    • run non-interactively
    • reproduce AGS 'as any student'
    • check environment
    • this gem: "For detailed usage, read the src code. The learning curve requires: Docker, Makefile, AWS Assume Role (to auth aws ecr and pull AGS image)

  • a submission tarball (submission.tar.gz) with required structure: -?

  • Maven

  • Java Grader java_grader.jar

  • Grading rubric example

  • aws command line interface

  • AGS Support for a task (what is task vs. module?) needs to be granted by TPZ staff:

  • a grader jarfile in target/grader.jar

  • access to the Cloud Computing Docker Registry (or our own Docker Registry?)

  • signature

  • semester

  • courseID <- TPZ needs to know about this already, assuming is 11791

  • upload-grader-plugin

Tricks

Cloud Computing course uses a Docker repository name like 939223853384.dkr.ecr.us-east-1.amazonaws.com (prepending an image name like cmucc/p0lg:v1)

To get Markdown or YAML from a Project page on theproject.zone append to the URL ".md" or ".yaml" resulting in something that looks like yaml:

--- !Manifest
course: <course-id>
sections:
  - !Section
    type: writeup
    id:
    file: writeup-<module-id>/000_introduction_to_big_data_analytics.md
    title: Introduction to Big Data Analytics
  - !Section
    type: assessme
    id: 2204 # the AssessMe id
    file:
    title:
  - !Section
    type: writeup
    id:
    file: writeup-<module-id>/010_aws_emr_azure_hdinsight_and_gcp_dataproc.md
    title: AWS EMR, Azure HDInsight and GCP Dataproc
  - !Section
    type: assessme
    id: 2205
    file:
    title:
  - !Section
    type: writeup
    id:
    file: writeup-<module-id>/020_big_data_processing_and_analysis.md
    title: Big Data Processing and Analysis
  - !Section
    type: assessme
    id: 2206
    file:
    title:
  - !Section
    type: writeup
    id:
    file: writeup-<module-id>/030_project_reflection_task_mandatory_graded.md
    title: Project Reflection Task (Mandatory, graded)
  - !Section
    type: writeup
    id:
    file: writeup-<module-id>/040_project_feedback.md
    title: Project Feedback
  - !Section
    type: writeup
    id:
    file: writeup-<module-id>/050_project_discussion.md
    title: Project Discussion

module: <module-id>

About

Notebook for ramping up on ACAI project