feat: export client-output-buffer-limit

[jdheyburn]

Fixes #932

Exports configured client-output-buffer-limit as a metric. Combined with mem_total_replication_buffers it will allow you to alert on when replication buffers are reaching the limit.

(mem_total_replication_buffers / config_client_output_buffer_limit_bytes{class="slave", limit="hard"}) > 0.7

Happy for suggestions on alternative metric name and labelling tactic.

[coveralls]

Pull Request Test Coverage Report for Build 391

Details

• 16 of 16 (100.0%) changed or added relevant lines in 1 file are covered.
• No unchanged relevant lines lost coverage.
• Overall coverage increased (+0.1%) to 84.02%

---

Totals
Change from base Build 387:	0.1%
Covered Lines:	1998
Relevant Lines:	2378

---

💛 - Coveralls

[oliver006]

Thanks for the PR - looks good!

[nantiferov]

@jdheyburn thank you for PR, I added mem_total_replication_buffers in 1.60, but haven't thought about collecting buffer limits.

P.S. Tested it quickly now, I think query for alert should look like this because of labels mismatch on left and right of /:

redis_mem_total_replication_buffers_bytes / sum(redis_config_client_output_buffer_limit_bytes{class="slave", limit="hard"}) without (class, limit)

[nantiferov]

Btw, @oliver006 I had similar case with sysctl values and node_exporter and they have similar functionality to config metrics, but it's possible to put numeric values into metric value and also generate metric name based on value labels.

https://github.com/prometheus/node_exporter/tree/master?tab=readme-ov-file#sysctl-collector

I think you might consider this as an option if someone will have time to improve config collector.