Charles Lowell
bde1631c83
When an error occurs in the indexer pipeline, it logs it out, and then drops it on the floor. This is problematic because externally, from the task scheduler's perspective, it appears to have succeeded, which is not true. As a result, these failures are invisible to tooling which introspects over tasks. This makes the index task promise resolve when the pipeline finishes as before, but it rejects if the pipeline failed. Signed-off-by: Charles Lowell <cowboyd@frontside.com>
search-backend-node
This plugin is part of a suite of plugins that comprise the Backstage search platform. This particular plugin is responsible for all aspects of the search indexing process, including:
- Providing connections to search engines where actual document indices live and queries can be made.
- Defining a mechanism for plugins to expose documents that they'd like to be
indexed (called
collators). - Defining a mechanism for plugins to add extra metadata to documents that the
source plugin may not be aware of (known as
decorators). - A scheduler that, at configurable intervals, compiles documents to be indexed and passes them to a search engine for indexing.
- A builder class to wire up all of the above.
- Naturally, types for all of the above.
Documentation on how to develop and improve the search platform is currently
centralized in the search plugin README.md.
For a better overview of how the search platform is put together, check the Backstage Search Architecture documentation.