Redis Cluster on GCP Doesn't Has Separate Resources Utilizations Metrics

[iqbaltkj]

Describe the problem
I run a Redis cluster with 5 shards and 2 replicas on GCP. Since the cluster has only one endpoint, so I added this Redis endpoint to the Redis exporter. I want to monitor the resource utilization of all my Redis instances, but the exporter only shows the total resources for all instances combined.

What version of redis_exporter are you running?
1.67

Running the exporter
Output metrics :

# HELP process_cpu_seconds_total Total user and system CPU time spent in seconds.
# TYPE process_cpu_seconds_total counter
process_cpu_seconds_total 950.44

I want to get something like this metrics output :

# HELP process_cpu_seconds_total Total user and system CPU time spent in seconds.
# TYPE process_cpu_seconds_total counter
process_cpu_seconds_total_run_id_$(n) 50.44
process_cpu_seconds_total_run_id_$(n) 80.44

Screenshots

Example Dashboard with the same issue

Additional context

1. Can the current exporter separate the resource utilization for each instance?
2. Can it work as a new feature or bug fix?

[oliver006]

Interesting question. Is there a way to get the CPU stats per process? Does "INFO" talk to a different process each time it's called? Is the process part of the response? If it is then it could be possible to add it to the metric (and others?) and record the metrics on a per-process basis.

[iqbaltkj]

I think each time it's called INFO its talk to a random different process.

[oliver006]

Ah interesting. Do you also see process_id? Maybe it's possible to add the ID (run_id or process_id) to certain process-level metrics to identify the individual processes that back the cloud service.

[iqbaltkj]

process_id is showing the same output each time it's called.

[oliver006]

Ah thanks, that actually makes sense. It's probably running in a container and so the pid is always 1.

"run_id" is prob better for identifying the different processes.

[iqbaltkj]

So is adding resources utilization or other metrics possible for every Redis instance while having only one endpoint?

[oliver006]

Sorry, only coming back to this now.
I think it would make sense to add a flag that would add the run_id to certain (process related) metrics. I don't have a ton of bandwidth for this right now (or a use case) but I can review a PR.