What was your favorite strategy if the Elasticsearch cluster failed or stopped responding? Did the search function on the site break, or was there a backup mechanism for direct queries to PostgreSQL?

In projects with Elasticsearch, fallback mechanisms are often implemented in case of cluster failure to prevent search functionality from breaking.

One strategy is to use fallback to PostgreSQL:

• When errors occur in Elasticsearch, requests are switched to the SQL database where the original data is stored.
• This can be implemented through a wrapper around the search service, which performs requests to PostgreSQL when ES errors occur.
• This approach reduces dependency on ES and increases fault tolerance.

Additionally:

• Monitor the cluster status and proactively switch traffic.
• Use caching of search results.
• Configure replicas and automatic recovery for ES.

An example of simplified logic in Go:

results, err := elasticsearchSearch(query)
if err != nil {
    results, err = postgresSearch(query)
    if err != nil {
        // handle error
    }
}