Fix cluster mode connecting to same node for all shards

[ajGingrich]

Summary

• Bug: In cluster mode, process_node() always connected to the config entry-point host/port (config.get("host") / config.get("port")) instead of the actual discovered node address passed via the node parameter. This meant every node in a Redis cluster was queried from the same instance, producing identical metrics across all masters and replicas.
• Fix: Use the discovered node's host and port (parsed from the node parameter) for the Redis connection, while preserving authentication and TLS settings from config.
• Scope: Only cluster mode was affected. In non-cluster (standalone) mode, process_database() constructs the node address from config["host"]:config["port"], so the values already matched.

Problem Detail

When process_database() discovers cluster nodes via CLUSTER NODES, it iterates over each discovered host:port and passes it to process_node(). However, process_node() was ignoring the node address and always creating the Redis client with:

client = get_redis_client(
    host=config.get("host"),     # <-- always the config entry-point
    port=int(config.get("port", 6379)),  # <-- always the config port
    ...
)

This caused all 6 rows in a 3-master/3-replica cluster to show identical values for memory, throughput, and command stats.

Changes

• osstats.py: Parse node into node_host and node_port, and use those for the get_redis_client() call
• test_osstats.py: Added test_process_node_connects_to_discovered_node which uses a different node address than the config entry-point, verifying get_redis_client is called with the discovered node's host/port

Test plan

• All 20 tests pass (pytest -v)
• Code formatting passes (black --check .)
• Re-run against a real Redis cluster to verify each node now reports distinct metrics

Made with Cursor