SlideRule

Building and Releasing SlideRule

4 minute read

The ICESat-2 SlideRule project consists of the following git repositories:

  • sliderule - server (open source, public)
  • sliderule-icesat2 - server plugin for icesat2 (open source, public)
  • sliderule-python - python client (open source, public)
  • sliderule-project - infrastructure, docker files, website, resources (private)

Releasing and deploying SlideRule consists of the following steps:

  1. Pre-release testing
  2. Release notes
  3. Tag release
  4. Build Docker images
  5. Build AMI
  6. Deploy cluster
  7. Post-release checkout
  8. Final steps

I. Pre-Release Testing

  1. SlideRule Self-Test 
    $ cd {root}/sliderule
$ make distclean
$ make development-config
$ make
$ sudo make install
$ sliderule scripts/tests/test_runner.lua
  2. ICESat-2 Plugin Self-Test 
    $ cd {root}/sliderule-icesat2
$ make distclean
$ make config
$ make
$ sudo make install
$ sliderule tests/test_runner.lua
  3. Python pytests locally (requires ATL03/06/08 files at /data) 
    $ cd {root}/sliderule-project
$ make monitor-docker
$ make monitor-docker-run # leave running
$ make sliderule-docker
$ make sliderule-docker-run # leave running
$ cd {root}/sliderule-python
$ conda activate sliderule # created from  {root}/sliderule-python/environment.yml
$ pytest --server="127.0.0.1" --asset="atlas-local"
  4. Python system tests locally (requires ATL03/06/08 files at /data) 
    $ cd {root}/sliderule-project
$ make sliderule-docker
$ make sliderule-docker-run # leave running
$ conda activate sliderule # created from  {root}/sliderule-python/environment.yml
$ python tests/algorithm.py
$ python tests/atl06gs.py
$ python tests/atl08class.py
  5. Deploy test cluster and run example workflows 
    $ cd {root}/sliderule-project
$ make sliderule-docker
$ docker push icesat2sliderule/sliderule:latest
#
# Log into AWS console and start a cloud shell
#
# create sliderule-node-latest AMI
# if it exists then delete ("deregister") AMI first via console
$ cd {root}/sliderule-project/packer
$ packer build sliderule-base.pkr.hcl
# wait until process completes and AMI image created
$ cd {root}/sliderule-project/terraform
$ terraform apply -var cluster_name=test
# type 'yes' at prompt after verifying that resources are prefixed with 'test'
# wait until process completes and nodes finish initializing
#
# Run Region of Interest on your local machine
#
$ cd {root}/sliderule-python
$ conda activate sliderule # created from  {root}/sliderule-python/environment.yml
$ python utils/region_of_interest.py {root}/sliderule-project/datasets/grandmesa.geojson {test-node-manager ip address}
#
# Start jupyter notebook and run through examples
#
#  - api_widgets_demo
#  - boulder_watershed_demo
#  - grand_mesa_demo
#  - single_track_demo
#
#
# Log into AWS console and start cloud shell
#
# delete test cluster
$ cd {root}/sliderule-project/terraform
$ terraform workspace select test
$ terraform destroy
# type 'yes' at prompt after verifying that resources are prefixed with 'test'
# wait until process completes and all resources are destroyed
  6. Run memory tests 
    #
# Run with Valigrind
#
$ cd {root}/sliderule-project
$ make distclean
$ make sliderule-config-valgrind
$ make sliderule
$ make sliderule-run-valgrind # leave running
$ cd {root}/sliderule-python
$ conda activate sliderule # created from  {root}/sliderule-python/environment.yml
$ python utils/region_of_interest.py {root}/sliderule-project/datasets/grandmesa.geojson
# there will be lots of timeouts; this is okay and tests error paths
# likely will have to stop the test early because it is taking too long (hours...)
#
# Run with Address Sanitizer
#
$ cd {root}/sliderule-project
$ make distclean
$ make sliderule-config-asan
$ make sliderule
$ make sliderule-run # leave running
$ cd {root}/sliderule-python
$ conda activate sliderule # created from  {root}/sliderule-python/environment.yml
$ python utils/region_of_interest.py {root}/sliderule-project/datasets/grandmesa.geojson
  7. Run H5Coro performance test 
    $ cd {root}/sliderule
$ make distclean
$ make python-config
$ make
$ cd build
$ cp {root}/sliderule-project/tests/perftest.py .
$ conda activate sliderule # created from  {root}/sliderule-python/environment.yml
$ python perftest.py
  8. Run static code analysis (scan-build) 
    $ cd {root}/sliderule
$ make distclean
$ make scan

II. Release Notes

Create a release note for the build in the sliderule-project repository, under jekyll/release_notes/, named release-v{x-y-z}.md.

III. Tag Release

  1. Tag the release in the git repositories 
    $ cd {root}/sliderule-project
$ make VERSION=x.y.z release-tag
  2. Publish the release on GitHub 
    # assumes that the github python environment exists, see release-prep target for details
$ cd {root}/sliderule-project
$ make VERSION=x.y.z release-github

IV. Build Docker Images

$ cd {root}/sliderule-project
$ make VERSION=x.y.z release-docker

V. Build AMI

#
# Log into AWS console and start a cloud shell
#
$ cd {root}/sliderule-project
$ make VERSION=x.y.z release-packer

VI. Deploy Cluster

#
# Log into AWS console and start a cloud shell
#
$ cd {root}/sliderule-project
$ make VERSION=x.y.z release-terraform

VII. Post-Release Checkout

  1. Run example workflows in python 
    $ cd {root}/sliderule-python
$ conda activate sliderule # created from  {root}/sliderule-python/environment.yml
$ python utils/query_version.py {vX.Y.Z-node-manager ip address}
$ python utils/query_services.py {vX.Y.Z-node-manager ip address}
$ python utils/region_of_interest.py {root}/sliderule-project/datasets/grandmesa.geojson {vX.Y.Z-node-manager ip address}
  2. Run pytest 
    $ cd {root}/sliderule-python
$ conda activate sliderule # created from  {root}/sliderule-python/environment.yml
$ pytest --server="{vX.Y.Z-node-manager ip address}"
  3. Check monitoring system at http://{vX.Y.Z-node-manager ip address}:3000/
    • View system metrics
    • View logs
    • View application metrics
  4. Check static website at http://{vX.Y.Z-node-manager ip address}/

VIII. Final Steps

  1. Route traffic from icesat2sliderule.org to the public IP address of the {vX.Y.Z-node-manager}
    • Go to “Route 53” in the AWS web console
    • Click on the “Hosted Zone” link under “DNS Management”
    • Click the icesat2sliderule.org domain link
    • Select the simple routing rule (type A) record
    • Click the button to “Edit record”
    • Change the Value of the IP address to the new IP address
    • Save changes
  2. Tear down previous deployment 
    #
# Log into AWS console and start a cloud shell
#
$ cd {root}/sliderule-project/terraform
$ terraform workspace select {previous release workspace}
$ terraform destroy

  3. Check CodeQL analysis (Actions –> CodeQL) on GitHub

  4. Check Conda-Forge deployment of Python client

Categories:

Updated: