llnl-sina 1.16.1

Contents

• Overview

• LC Setup

• Standard Setup

• Manual Setup

  • Creating the Environment

  • Installing Software Dependencies

• Activating and Deactivating the Virtual Environment

• Using the Example Notebooks and Datasets

• Testing

• Supported Environments

• Database Support

Overview

Sina’s Python component is a tool for making simulation (meta)data collection and exploration simple.

It works by collecting information from code runs, logs, and other outputs into a common file format which can then be passed off to one of Sina’s supported backends, all of which are queried using the same user-friendly Python API. To the end user, this means that important data can be accessed through Python scripts, GUIs (Jupyter notebooks) etc. with all the speed of a database and none of the complexity (the user never has to interact with database architecture), nor any of the traditional headaches of parsing logs or remembering which file contains what.

Sina is integrated into a number of LLNL physics codes to capture simulation data; look for the _sina.json! If your code isn’t configured to output Sina, but you’d like it to be, we may be able to work with the code team to integrate it–you can reach us at weave-support@llnl.gov, or check out the WEAVE project on Gitlab and Teams.

The instructions below will guide you through setting up a virtual environment for Sina (or installing it in one that already exists), running example notebooks, and getting dependencies for your backend(s) of choice. Note that SQL will always be available as the “default” backend. Once you’re done with setup, a quickstart tutorial can be found in notebook form at <sina_root>/examples/basic_usage.ipynb.

Remember that, if you’re on LC, each time you log in you’ll first need to activate the environment. When you’re done, we recommend you deactivate the virtual environment to get back to your default environment or end your session.

LC Setup

If you’re on an LC machine, you can use a virtual environment with dependencies already installed:

$ source /collab/usr/gapps/wf/releases/sina/bin/activate

The above is for bash; other activation scripts, e.g. activate.csh, can be found in the same directory.

Sina will now be available for use via Python virtual environment, and can be tested with sina -h (which should display a help message). When you’re done, use deactivate to exit the virtual environment. Note that this will be the release (master) Sina version–if you want to use Sina Develop, keep reading!

If you run into issues with the LC virtual environment, please email us at weave-support@llnl.gov.

Standard Non-LC Setup

Sina is available on PyPi:

$ pip install llnl-sina

However, this will only give you access to the release version! Non-release versions are not available externally. Internal users looking to use our development version, or wanting to contribute to Sina, clone us from CZ Gitlab. External contributors should clone us from the LLNL Github.

After cloning, install like any other standard python lib (cd to here, pip install .).

Notably, Sina contains a number of modules:

• jupyter adds support for jupyter notebook visualizations

• mysql adds support for using mysql and mariadb

• cassandra adds limited support for Cassandra (NoSQL), but is deprecated

• development adds dependencies for our test/CI

If you want any of the above, add them to your install command like:

$ pip install .[jupyter,development]

LC users have all optional dependencies available through the sina venv.

Using the Example Notebooks and Datasets

Sina contains tutorials in the form of Jupyter notebooks. Files are stored in the examples directory (found in the sina root folder alongside the python and cpp folders), and are organized by dataset, with data_overview.rst containing descriptions of each set. To use the notebooks, you’ll first need to run getting_started.ipynb (also in the examples directory) from the LC Jupyter server at lc.llnl.gov/jupyter. This will create a Jupyter kernel from your current virtual environment, making anything installed in it available to the notebook. After that, you’ll be ready to run the rest of the notebooks. If you’re not working on LC, you can also set Jupyter up locally: run make Jupyter from the python folder, then jupyter notebook. This will open a webpage similar to what you’d see accessing LC’s Jupyter server.

Most notebooks rely on sample datasets. Pre-built sets are deployed with Sina to the LC, but you can build them locally as well to experiment with Sina. Go into any dataset folder (the NOAA set is well-sized for experimentation) and ./build_db.sh. Note that you’ll need Sina available to do so, see the section on virtual environments.

To clean all output from the notebooks:

(venv) $ make clean-notebooks

Testing

Sina uses gitlab CI to test out MRs. To run a local equivalent, you can use pylint and the tests found in the scripts folder. For example:

$ python3 -m venv sample_venv && source sample_venv/bin/activate
$ pip install -e.[development]
$ pytest -v -m "not cassandra and not mysql"
$ ../scripts/test_*

This will install all necessary dependencies and run most tests. Excluded test categories (ex: mysql) require additional setup (ex: sacrificial database) and may be better left to the CI; if you want example setups, see the .gitlab-ci.yml.

If you need a dependency added, email us so I can re-run the nightly environment rebuild.

Supported Environments

Sina is most regularly tested against Python 3.9 and Python 3.14 in a RHEL4 environment.

Database Support

As mentioned above, “stock” Sina doesn’t include dependencies for databases beyond SQLite, but Sina supports and test against them. External users need to install them, ex: pip install .[mysql], and also need to have a reachable database.

LC venv users already have all additional modules installed, so specify a myql connector string in place of a db name and you should be good to go. If you used LaunchIT (lit.llnl.gov) to provision your MariaDB instance, they make this very easy to get, look for the “Sina Connection String” on your dashboard and paste it in like so:

$ sina ingest --database-type=sql --database "mysql+mysqlconnector://host:port/?read_default_file=~/.my.cnf"

LaunchIT also provides contents for the .my.cnf file, which you can paste to a location of your choosing, with the above assuming you placed it in $HOME.

Here’s how to use it with the API:

datastore = sina.connect("mysql+mysqlconnector://host:port/?read_default_file=~/.my.cnf")