redis slow log metrics

[Colstuwjx]

As redis has slow log metrics to show the slowest CMD, we should export these to prom. Thus, grafana could show the topK slow CMD, and makes redis ops feel happy.

[oliver006]

@Colstuwjx - what do you think would be useful to export for the slow log?

I could see the number of slow log entries being useful (e.g. export SLOWLOG LEN) but the slow log contents itself seems not really suitable for exporting.

Thoughts?

[Colstuwjx]

Yes, I agree that. Instead collecting the detail slow log CMD as "metrics" into prometheus, I'd like to scrape them into Elastic.

Thanks a lot for your reply.

[mazzy89]

i'd need a slow queries view in Grafana maybe showing the last 10 slow query. My idea would be to have a metric that would look like this:

redis_slowlog_info{command='', id='', time=''} <execution_duration>

where the values are the result of the SLOWLOG query properly parsed.

wouldn't be hard to add this to the exporter. What do you think @oliver006 ? I'd prepare a PR for today

[oliver006]

I think that's something for log collection, not for metrics and therefor out of scope for the exporter.

[siwyd]

FYI, you can use Filebeat to export slow logs: https://www.elastic.co/blog/brewing-in-beats-read-redis-slow-logs-with-filebeat

[madhur]

I think slow queries logs are really needed to be exported for prometheus. Atleast the top 10