RecycledCloud/smartctl_exporter
6
0
Fork 0
mirror of https://github.com/prometheus-community/smartctl_exporter.git synced 2024-12-21 02:21:55 +01:00
Code Issues Projects Releases 1 Packages Wiki Activity

fix: correct smartctl_device_bytes_written & smartctl_device_bytes_read for NVMe (#211)

Browse source 
The NVMe specification says that the controller is responsible for
reporting "Data Units Read" & "Data Units Written" converted as needed
for logicial block sizes other than 512-bytes. smartmontools already has
the correct behavior.

What is correct in this case? For now, track what smartmontools does:
take the counter, multiply by 512*1000, report the value.

We should be clear that it means the drive has read/written at most
that many bytes.

This has a few impacts:
- NVME devices will now show these metrics, if they did not before.
- NVME devices with blocksize other than 512-bytes may have previously
  reported inflated metrics, but are now corrected (is this worthy of
  larger notice in changelogs?)

Reference: 11415ee0b9/smartmontools/nvmeprint.cpp (L394-L397)
Closes: https://github.com/prometheus-community/smartctl_exporter/issues/122

Signed-off-by: Robin H. Johnson <rjohnson@coreweave.com>
This commit is contained in:
Robin H. Johnson 2024-04-09 00:40:44 -07:00 committed by GitHub
parent d6cd1815eb
commit ea8a38384b
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
1 changed files with 33 additions and 10 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

43
smartctl.go
View file

@ -370,34 +370,57 @@ func (smart *SMARTctl) mineNvmeNumErrLogEntries() {
)
}
// https://nvmexpress.org/wp-content/uploads/NVM-Express-NVM-Command-Set-Specification-1.0d-2023.12.28-Ratified.pdf
// 4.1.4.2 SMART / Health Information (02h)
// The SMART / Health Information log page is as defined in the NVM Express Base Specification. For the
// Data Units Read and Data Units Written fields, when the logical block size is a value other than 512 bytes,
// the controller shall convert the amount of data read to 512 byte units.
// https://nvmexpress.org/wp-content/uploads/NVM-Express-Base-Specification-2.0d-2024.01.11-Ratified.pdf
// Figure 208: SMART / Health Information Log Page
// Bytes 47:32
// Data Units Read: Contains the number of 512 byte data units the host has read from the
// controller as part of processing a SMART Data Units Read Command; this value does not
// include metadata. This value is reported in thousands (i.e., a value of 1 corresponds to 1,000
// units of 512 bytes read) and is rounded up (e.g., one indicates that the number of 512 byte
// data units read is from 1 to 1,000, three indicates that the number of 512 byte data units read
// is from 2,001 to 3,000).
//
// A value of 0h in this field indicates that the number of SMART Data Units Read is not reported.
//
// Bytes 63:48
//
// Data Units Written: Contains the number of 512 byte data units the host has written to the ...
// (the same as Data Units Read)
func (smart *SMARTctl) mineNvmeBytesRead() {
blockSize := smart.json.Get("logical_block_size")
data_units_read := smart.json.Get("nvme_smart_health_information_log.data_units_read")
if !blockSize.Exists() || !data_units_read.Exists() {
// 0 => not reported by underlying hardware
if !data_units_read.Exists() || data_units_read.Int() == 0 {
return
}
smart.ch <- prometheus.MustNewConstMetric(
metricDeviceBytesRead,
prometheus.CounterValue,
// This value is reported in thousands (i.e., a value of 1 corresponds to 1000 units of 512 bytes written) and is rounded up.
// When the LBA size is a value other than 512 bytes, the controller shall convert the amount of data written to 512 byte units.
data_units_read.Float()*1000.0*blockSize.Float(),
// WARNING: Float64 will lose precision when drives reach ~32EiB read/write
// The underlying data_units_written,data_units_read are 128-bit integers
data_units_read.Float()*1000.0*512.0,
smart.device.device,
)
}
func (smart *SMARTctl) mineNvmeBytesWritten() {
blockSize := smart.json.Get("logical_block_size")
data_units_written := smart.json.Get("nvme_smart_health_information_log.data_units_written")
if !blockSize.Exists() || !data_units_written.Exists() {
// 0 => not reported by underlying hardware
if !data_units_written.Exists() || data_units_written.Int() == 0 {
return
}
smart.ch <- prometheus.MustNewConstMetric(
metricDeviceBytesWritten,
prometheus.CounterValue,
// This value is reported in thousands (i.e., a value of 1 corresponds to 1000 units of 512 bytes written) and is rounded up.
// When the LBA size is a value other than 512 bytes, the controller shall convert the amount of data written to 512 byte units.
data_units_written.Float()*1000.0*blockSize.Float(),
// WARNING: Float64 will lose precision when drives reach ~32EiB read/write
// The underlying data_units_written,data_units_read are 128-bit integers
data_units_written.Float()*1000.0*512.0,
smart.device.device,
)
}