elasticsearch-templates

The repository contains scripts and sources to generate Elasticsearch templates that comply with Common Data Model.

Problem Statement

We are trying to solve the problem of conflicts and inconsistencies in log data as collected by, and from, different subsystems stored together as a unified data set under one warehouse.

Namespace hierarchy

Namespace hierarchy on the log metadata is the key concept. We use the Elasticsearch index templates and document mapping to cast the common metadata keys into usable documents.

Namespace corresponds to a top-level JSON key of Elasticsearch document. Namespace is usually defined per individual app or subsystem, so that different applications/subsystems not conflict in various metadata fields.

Adding new namespace

Create a namespace definition file in namespaces/ folder.

Adding new Elasticsearch template

Create a sub-folder in templates/ folder, named as the desired template. (Alternatively copy/modify one of the existing template folders)

Add/modify template.yml definition file to include proper namespace definitions. See for example templates/openshift/README.md for the details.

Elasticsearch versions support

Support for multiple Elasticsearch versions has been added. Resulting files (ie. index-templates or index-patterns) are generated for each supported version of Elasticsearch. Target version of ES is encoded into the file name.

List of currently supported ES versions can be find in scripts/supported_versions.py.

The idea is that all the input file templates and data are formatted according to the latest supported ES version and scripts handle backward data and format conversions for older ES versions. As part of unit testing the generated data is compared to released common data files (automatically downloaded from GitHub during tests).

Generating documentation

Use the makefile in the templates/ folder.

Alternatively, run the following command: python ./scripts/generate_template.py (path to template in templates/) namespaces/ --doc.

The generated file looks like "xxx.asciidoc".

Viewing the documentation

Install the asciidoc viewer in web browser.

Open the local path to the asciidoc file "xxx.asciidoc" in your browser.

Releasing a new version of the data model

First, generate index templates (for Elasticsearch) and index patterns (for Kibana).

$ cd <project_root>
$ make clean
$ make

Create a new release tag in repo and push it into GitHub.

$ git tag -a 0.0.24 -m "Release 0.0.24"

# We can check the tag is attached to the latest commit now 
$ git log --oneline -n 2
c16dc2c (HEAD -> master, tag: 0.0.24, origin/master, origin/HEAD) Fix index patterns
39d0b71 (tag: 0.0.23) Update model & Bump to 2020.01.23

# Push tag into remote GitHub repo
$ git push origin --tags
Total 0 (delta 0), reused 0 (delta 0)
To github.com:ViaQ/elasticsearch-templates.git
 * [new tag]         0.0.24 -> 0.0.24 

Create a new release in GitHub project release.

• create a new release draft from newly created tag
• provide meaningful description and manually attach files belonging to the release
  • Usually the list of the files is the same as in the previous release except when it is not :-) (i.e. if there is any significant change)
• publish the release

Once a new release is published you can use update-viaq-data-model.sh script to pull released files into AOL and prepare a new PR with updated data model.