「PVE」- Metric Server，指标含义

官方文档并未给出 Metric 的含义（03/09/2023 PVE 7.3），我们需要通过其他手段来获取相关信息：
1）分析 pvestatd 源码：git.proxmox.com Git – pve-manager.git/blob – PVE/Service/pvestatd.pm

指标分类

通过 Flux 查询，确定其中包含的指标（Measurement）： 

// -------------------------------------------------------- // 查看全部 Measurement 指标；

import "influxdata/influxdb/schema"
schema.measurements(bucket: "pve-metric-server")

ballooninfo ----------------------------------------------- // 包含 Host 指标，部分指标与 Guest 相关；
proxmox-support ------------------------------------------- // 包含 Host 指标，部分指标与 Guest 相关；
system ---------------------------------------------------- // 包含 Host 指标，部分指标与 Guest 相关；
blockstat ------------------------------------------------- // 包含 Host 指标，部分指标与 Guest 相关；
nics ------------------------------------------------------ // 包含 Host 指标，部分指标与 Guest 相关；
cpustat --------------------------------------------------- // 仅包含 Host 指标；
memory ---------------------------------------------------- // 仅包含 Host 指标；

// -------------------------------------------------------- // 

import "influxdata/influxdb/schema"
schema.measurementTagKeys(
    bucket: "pve-metric-server",
    measurement: "proxmox-support",
)


import "influxdata/influxdb/schema"

schema.measurementTagValues(
    bucket: "pve-metric-server",
    measurement: "ballooninfo",
    tag: "host",
)

Tags

_start
_stop
_field
_measurement
host
instance
nodename
object
type
vmid

host: pve-node-name pve-guest-name
nodename: pve-node-name
object: nodes qemu storages …
type: dir lvm lvmthin nfs …

Fields

active，actual，avail，avg1，avg15，avg5，balloon，bavail，bfree，blocks，content，cpu，cpus，ctime，disk，diskread，diskwrite，enabled，failed_flush_operations，failed_rd_operations，failed_unmap_operations，failed_wr_operations，favail，ffree，files，flush_operations，flush_total_time_ns，fper，free_mem，freemem，fused，guest，guest_nice，idle，idle_time_ns，invalid_flush_operations，invalid_rd_operations，invalid_unmap_operations，invalid_wr_operations，iowait，irq，last_update，major_page_faults，max_mem，maxdisk，maxmem，mem，mem_swapped_in，mem_swapped_out，memfree，memshared，memtotal，memused，minor_page_faults，name，netin，netout，nice，pbs-library-version，per，pid，qmpstatus，rd_bytes，rd_merged，rd_operations，rd_total_time_ns，receive，running-machine，running-qemu，serial，shared，softirq，status，steal，su_bavail，su_blocks，su_favail，su_files，sum，swapfree，swaptotal，swapused，system，template，total，total_mem，transmit，type，unmap_bytes，unmap_merged，unmap_operations，unmap_total_time_ns，uptime，used，user，user_bavail，user_blocks，user_favail，user_files，user_fused，user_used，wait，wr_bytes，wr_highest_offset，wr_merged，wr_operations，wr_total_time_ns，

ballooninfo

Field

// -------------------------------------------------------- // 查看该 Measurement 下全部 Field 信息；

import "influxdata/influxdb/schema"
schema.measurementFieldKeys(
    bucket: "pve-metric-server",
    measurement: "ballooninfo",
)

actual，free_mem，last_update，major_page_faults，max_mem，mem_swapped_in，mem_swapped_out，minor_page_faults，total_mem

proxmox-support

Field

import "influxdata/influxdb/schema"
schema.measurementFieldKeys(
    bucket: "pve-metric-server",
    measurement: "proxmox-support",
)

pbs-library-version：该 measurement 中，仅包含 pbs-library-version 字段。鉴于其并不包含其他指标数据，所以我们暂时忽略；

system

Field

import "influxdata/influxdb/schema"
schema.measurementFieldKeys(
    bucket: "pve-metric-server",
    measurement: "system",
)

active，avail，balloon，content，cpu，cpus，disk，diskread，diskwrite，enabled，freemem，maxdisk，maxmem，mem，name，netin，netout，pid，qmpstatus，running-machine，running-qemu，serial，shared，status，template，total，type，uptime，used，

disk，diskread，diskwrite，maxdisk：该指标仅与 Guest 相关；

Query

from(bucket: "pve-metric-server")
    |> range(start: v.timeRangeStart, stop: v.timeRangeStop)
    |> filter(fn: (r) => r["_measurement"] == "system")
    |> filter(fn: (r) => r._field == "disk" or r._field == "maxdisk")
    |> filter(fn: (r) => r.disk != 0)

blockstat（host-and-guest-related）

fields

import "influxdata/influxdb/schema"
schema.measurementFieldKeys(
    bucket: "pve-metric-server",
    measurement: "blockstat",
)

bavail，bfree，blocks，failed_flush_operations，failed_rd_operations，failed_unmap_operations，failed_wr_operations，favail，ffree，files，flush_operations，flush_total_time_ns，fper，fused，idle_time_ns，invalid_flush_operations，invalid_rd_operations，invalid_unmap_operations，invalid_wr_operations，per，rd_bytes，rd_merged，rd_operations，rd_total_time_ns，su_bavail，su_blocks，su_favail，su_files，unmap_bytes，unmap_merged，unmap_operations，unmap_total_time_ns，used，user_bavail，user_blocks，user_favail，user_files，user_fused，user_used，wr_bytes，wr_highest_offset，wr_merged，wr_operations，wr_total_time_ns，

host-related fields

获取与 PVE Node 相关的指标： 

from(bucket: "pve-metric-server")
    |> range(start: v.timeRangeStart, stop: v.timeRangeStop)
    |> filter(fn: (r) => r["_measurement"] == "blockstat" and r.object == "nodes")
    |> keep(columns: ["_field"])
    |> distinct(column: "_field")

bavail，bfree，blocks，favail，ffree，files，fper，fused，per，su_bavail，su_blocks，su_favail，su_files，used，user_bavail，user_blocks，user_favail，user_files，user_fused，user_used，

PVE 使用 Perl 语言，通过 Filesys::Df 获取磁盘信息，所以这些指标也是 df 相关的输出。参考 Filesys::Df 文档，以获取相关字段含义；

The keys available in the hash are as follows:
{blocks} = Total blocks on the filesystem.
{bfree} = Total blocks free on the filesystem.
{bavail} = Total blocks available to the user executing the Perl application. This can be different than {bfree} if you have per-user quotas on the filesystem, or if the super user has a reserved amount. {bavail} can also be a negative value because of this. For instance if there is more space being used then you have available to you.
{used} = Total blocks used on the filesystem.
{per} = Percent of disk space used. This is based on the disk space available to the user executing the application. In other words, if the filesystem has 10% of its space reserved for the superuser, then the percent used can go up to 110%.

Here are the available inode keys:
{files} = Total inodes on the filesystem.
{ffree} = Total inodes free on the filesystem.
{favail} = Total inodes available to the user executing the application. See the rules for the {bavail} key.
{fused} = Total inodes used on the filesystem.
{fper} = Percent of inodes used on the filesystem. See rules for the {per} key.

guest-related fields

虚拟机实例相关： 

from(bucket: "pve-metric-server")
    |> range(start: v.timeRangeStart, stop: v.timeRangeStop)
    |> filter(fn: (r) => r["_measurement"] == "blockstat")
    |> filter(fn: (r) => r.host == "k8s130-wn130-206" and r.instance == "scsi0")

nics

PVE Node： 

from(bucket: "pve-metric-server")
    |> range(start: v.timeRangeStart, stop: v.timeRangeStop)
    |> filter(fn: (r) => r["_measurement"] == "nics")
    |> filter(fn: (r) => r.object == "nodes" and r.host == "${server}")

Field

import "influxdata/influxdb/schema"
schema.measurementFieldKeys(
    bucket: "pve-metric-server",
    measurement: "nics",
)

netin、netout => 其为与 Guest 相关的指标；
receive、transmit => 其为与 Host 相关的指标； 以 Byte 为单位；

cpustat（host-related）

Field

import "influxdata/influxdb/schema"
schema.measurementFieldKeys(
    bucket: "pve-metric-server",
    measurement: "cpustat",
)

avg1，avg15，avg5，cpu，cpus，ctime，guest，guest_nice，idle，iowait，irq，nice，softirq，steal，sum，system，total，used，user，wait，

判断 Host/Guest 相关： 

from(bucket: "pve-metric-server")
  |> range(start: v.timeRangeStart, stop: v.timeRangeStop)
  |> filter(fn: (r) => r["_measurement"] == "cpustat")
  |> drop(columns: ["_field"])
  |> distinct(column: "host")

memory（host-related）

Field

import "influxdata/influxdb/schema"
schema.measurementFieldKeys(
    bucket: "pve-metric-server",
    measurement: "memory",
)

memfree，memshared，memtotal，memused，swapfree，swaptotal，swapused，

判断 Host/Guest 相关： 

from(bucket: "pve-metric-server")
  |> range(start: v.timeRangeStart, stop: v.timeRangeStop)
  |> filter(fn: (r) => r["_measurement"] == "memory")
  |> drop(columns: ["_field"])
  |> distinct(column: "host")

Filed under: K4NZDROID - @ 12:21 AM