Redis Cluster exporter shows identical key values across nodes — intended behavior or bug?

[abregar]

Describe the problem
Just recently set up Prometheus monitoring for a Redis Cluster (3 masters, 3 replicas) using this redis_exporter. Everything looked fine at first, metrics were scraping correctly, but I noticed that all nodes were exposing identical key values. To test further, I intentionally stopped one master instance, yet the exporter still reported redis_up = 1 for that node. From what I can tell, redis_up seems to reflect process reachability (TCP/INFO) rather than actual cluster health, which makes it confusing when a node is flagged as failed in CLUSTER NODES but still shows up as healthy in Prometheus. I found a few potentially similar reports in one of the comments - #979 (comment) which identifies similar behavior.

Running the exporter
--web.listen-address=127.0.0.1:9121 --is-cluster=true

Screenshots
Here's an example of redis_instance_info showing the case further

Additional context
Add any other context about the question here.

[oliver006]

Can you provide more information on your setup?
Are you connecting always to the same host or are you connecting to all nodes in your cluster?

[abregar]

Setup: two hosts, each running three Redis 8.2 cluster instances (3 masters, 3 replicas; ports increment by 1) with replicas anti-affined across hosts. One redis_exporter per host using HTTP SD; http://:9121/discover-cluster-nodes reliably lists all six nodes. Prometheus scrapes via http_sd_configs, i.e., it connects only to the two exporters (no direct per-instance scraping). Despite per-instance series, key metrics show identical values across nodes; it appears each exporter may proxy primarily through the first local instance. Cause unknown—seeking guidance or confirmations.