Axes and labelling for disk I/O graphs about grafana-dashboards

[lethargosapatheia]

@candlerb As I don't have a medium account, I'd like to thank you here :) Really helped me out especially in regards to r_await and w_await - I was using them wrongly without the "completed" metrics. The io depth explanation was also very insightful.

[candlerb]

This might be a bit tricky to do as I'm on a newer version of grafana (7.3.7) than the one in the repo (6.7.3) - and the export includes quite a lot of differences due to Grafana adding new fields.

Anyway, I've put my version up temporarily at https://psg.com/~brian/ne1860-1611854721829.json

The main changes I made were:

• set the axis labels and units for queue size and time spent doing I/O
• fix the disk read/write time queries to match what iostat does to generate r_await and w_await: e.g. irate(node_disk_read_time_seconds_total{instance="$node",job="$job"}[5m]) / irate(node_disk_reads_completed_total{instance="$node",job="$job"}[5m])
• change to SI units on summary "I/O Usage Read / Write" graph to be consistent with "Disk R/W Data"
• add descriptions to all the detail graphs based on the iostat(1) manpage

Adding average request size graphs (rareq-sz and wareq-sz) might be useful too, but I didn't do that.

[rfrail3]

Brian, excellent article, in depth and concise research, it was a pleasure to see all that information in one place, congratulations.

I just upload an update revision of the dashboard with Grafana 7.3.7.

Please, feel free to send a PR with the updates that you consider opportune, it'll be wellcome.

[rfrail3]

Brian, It's difficult to get a clear diff with the changes between your version and the current one. On commit a154f92 I upload the ones that I found, I think that these are all and it's all done.