Pure Storage FlashBlade system hardware dashboard health completely broken

[florianmulatz]

Shows "No data" or "N/A" on any graph / stat-panel

[gloacai]

This looks like this dashboard is a copy and paste of the flash array one. All the metrics it's looking for are flasharray ones (purefa_ vice purefb_). Digging at it more in current master branch flashblade_collector/flashblade_metrics/array_hardware_metrics.py needs a lot of love compared to the flasharray version since most of the metrics this dashboard is looking for isn't being exported yet.

[gloacai]

As said above, in general need to rework flashblade_collector/flashblade_metrics/array_hardware_metrics.py to expose more metrics.

Testing against Purity//FB3.1.4 (chassis, blades, power supplies, fabric modules) and reading https://purity-fb.readthedocs.io/en/latest/Hardware/

• array name, firmware version, average iops, average IO latency, and average throughput: fixed by swapping from purefa_* to purefb_*
• average power voltage: There doesn't appear to be a way to query for voltage used so that panel will need to be deleted
• average temperature: All the components have a temperature field in their rest api but their value is None for all of them, and looking at the Pure's own management web tool I don't see temperature reported anywhere. Guessing this panel should be deleted for now
• location: delete this panel as the site is not a field reported by the pure in any query
• capacity usage: need to change the query to sum(purefb_array_space_bytes{job=~"$job",instance="$instance",dimension="total_physical"}) / sum(purefb_array_capacity_bytes{job=~"$job",instance="$instance"})
• data reduction: need to change the query to sum(purefb_array_space_data_reduction{job=~"$job",instance="$instance"})
• system events: the purefb_alerts_total doesn't exist but it looks like it can be queried https://purity-fb.readthedocs.io/en/latest/AlertsApi/#list_alerts
• nvram*: there is not a hardware component type for nvram for the blade, need to delete these panels
• disk bays: not applicable to a FlashBlade array, should swap to monitoring the fb component type instead
• failed SSDs: doesn't see like the API exposes down to that detail
• Controller should be swapped to fabric modules and the sub components are different

[sdodsley]

Resolved by #43