Redis v5 gem feasibility notes

As redis-rb v5 separately packages Redis Cluster functionalities in redis-clustering gem that uses redis-cluster-client, we would need to examine and verify that there will be no regression when upgrading from v4.8.0 to v5.0.8.

This issue tracks any new findings that require a patch or upstreamed fix.

Important recent fixes

• handle redirection for transactions
• transaction support
• auto-reconnecting after fork

Performance changes

• start-up tuning which allows users to configure slow-command-specific timeouts
• add an environment variable to be able to configure a data type of slots and nodes mapping

On concurrency

Since redis-cluster-client provides an option to run all pipelines on the main thread, we will transit into that for a start. This ensures that there are minimal behavioural changes when upgrading the gem.

The option to use concurrent workers can be a way to reduce the high latency of tail events in #2503 (moved). It can be explored separately from this epic.

Note: v5.0.8 released in https://github.com/redis/redis-rb/commit/2b183addb9175641a48e1d8616772282a31c0532 which contains several key changes

• transaction support -- https://github.com/redis/redis-rb/pull/1226
• compatibility with redis-cluster-client >= v6.0.0 -- https://github.com/redis/redis-rb/pull/1222