Addition fs metrics

[ebpitts]

I see es-stats collecting the total_in_bytes_and available_in_bytes stats. The Cluster Stats API makes additional information about i/o available, such as

• disk_reads
• disk_writes
• disk_io_op (sum of disk_reads and disk_writes)
• disk_read_size_in_bytes
• disk_write_size_in_bytes
• disk_io_size_in_bytes (sum of disk_read_size_in_bytes and disk_write_size_in_bytes)

Have you considered collecting them?

[jamiealquiza]

I actually have, but see it as ultimately a better metric to collect externally (eg via collectd at each node) given per-node granularity and control over resolution. However, collection from es-stats is very inexpensive and do not see a reason it shouldn't be added anyway.