Stabilize the project website pipeline and website search infrastructure

[zelinh]

Currently the infrastructure of building, deploying & indexing our project & documentation websites is consistently failing.

The project is managed by codePipeline; when build completes, indexing function will start and ingest the data source to our cluster spun up manually from tarball.

This process would often fail. When indexing function tries to index the datasource to bulk the new index into the cluster, it also removed the previous index. However, when multiple workflow run at the same time, if the cluster has already removed one shard and another workflow is trying to remove the same shard, it would halt and eventually fail.

We are planning to create new pipelines for building project website and documentation websites.

We will stabilize the infra side for the website search feature and the way we indexing all the pages.

Acceptance criteria:

• The pipeline for building, deploying and indexing the project & documentation websites passes successfully.
• Combine multiple pipelines for each versions into a single pipeline to avoid race condition.
• Split different versions of documentation into separate individual index for less duplicate search results.