Disk I/O metrics should not report values for multiple partitions on the same device

[stefanotorresi]

I think I/O metrics should only be reported per device, not per partition.

e.g.

hanadb_disk_io_latency_ms{disk="vda",host="stefanotorresi-hana01"} 0.58
hanadb_disk_io_latency_ms{disk="vda1",host="stefanotorresi-hana01"} 0.0
hanadb_disk_io_latency_ms{disk="vda2",host="stefanotorresi-hana01"} 0.0
hanadb_disk_io_latency_ms{disk="vda3",host="stefanotorresi-hana01"} 0.6
hanadb_disk_io_latency_ms{disk="vdb",host="stefanotorresi-hana01"} 0.28
hanadb_disk_io_latency_ms{disk="vdb1",host="stefanotorresi-hana01"} 0.28
hanadb_disk_io_latency_ms{disk="vdc",host="stefanotorresi-hana01"} 0.35

only vda vdb and vdc lines should be reported, not the single partitions vda1 vda2 vda3 vdb1, which often are duplicated, wrong or slightly skewed values.

[diegoakechi]

@stefanotorresi This information is extracted by this query and comes from the internal HANA monitoring views:

SELECT host, disk, queue_length, srv_ms + wait_ms latency_ms, srv_ms, wait_ms, io_per_s, tp_kbps FROM( SELECT MAX(TIMESTAMP) timestamp, host, measured_element_name disk, MAX(MAP(caption, 'Queue Length', TO_NUMBER(value), 0)) queue_length, MAX(MAP(caption, 'Service Time', TO_NUMBER(value), 0)) srv_ms, MAX(MAP(caption, 'Wait Time', TO_NUMBER(value), 0)) wait_ms, MAX(MAP(caption, 'I/O Rate', TO_NUMBER(value), 0)) io_per_s, MAX(MAP(caption, 'Total Throughput', TO_NUMBER(value), 0)) tp_kbps FROM sys.m_host_agent_metrics WHERE measured_element_type = 'Disk' GROUP BY host, measured_element_name);

And HANA itself is storing the info on this granularity, so may be there is a good reason for that.

So, I would recommend some investigation with SAP about that. For the dashboard, I wouldn't be against summarizing it on the graphs... What do you think?