Worthless disk metrics

[jwwalton]

not sure what i'm supposed to do with these:

node_disk_bytes_read{device="dm-0"} 4.14021632e+08
node_disk_bytes_read{device="dm-1"} 1.6781312e+07
node_disk_bytes_read{device="dm-10"} 5.364992e+08
node_disk_bytes_read{device="dm-11"} 4.268032e+06
node_disk_bytes_read{device="dm-12"} 3.3796864e+08
node_disk_bytes_read{device="dm-13"} 3.3776384e+08
node_disk_bytes_read{device="dm-14"} 8.690176e+06
node_disk_bytes_read{device="dm-15"} 3.74784e+06
node_disk_bytes_read{device="dm-16"} 2.64704e+06
node_disk_bytes_read{device="dm-17"} 2.121728e+06
node_disk_bytes_read{device="dm-2"} 1.9963904e+07
node_disk_bytes_read{device="dm-3"} 1.0097152e+08
node_disk_bytes_read{device="dm-4"} 3.63190272e+08
node_disk_bytes_read{device="dm-5"} 5.328896e+06
node_disk_bytes_read{device="dm-6"} 3.38434048e+08
node_disk_bytes_read{device="dm-7"} 2.23399424e+09
node_disk_bytes_read{device="dm-8"} 1.08535808e+09
node_disk_bytes_read{device="dm-9"} 6.07663104e+08
node_disk_bytes_read{device="sda"} 2.674330624e+09
node_disk_bytes_read{device="sdb"} 4.9864192e+08
node_disk_bytes_read{device="sdc"} 9.11872e+06

these device names are not stable at all. so the metrics are mostly worthless unless you reset them at every boot.

[SuperQ]

For usage question, please use our community contacts.

[jwwalton]

i don't have any questions about usage. it just means that node exporters disk metrics are pretty much worthless for most people that actually reboot their systems once in a while. collectd and every other thing under the sun solves this problem by grabbing info from udev. so consider this a feature request for udev support instead.

[joehandzik]

@jwwalton Take a look at issue #304. I see your point (it's an issue we deal with in the context of Lustre, for example), but it seems like there are some workarounds available.