Prometheus Node_exporter 之 Memory Detail Vmstat

Memory Detail Vmstat

查看/proc/vmstat 文件的内容

1. Memory Pages In / Out

type: Graph
Unit: short
Label: Pages
Pagesin - 从启动到现在读入的内存页数的速率(5分钟内) /proc/vmstat pgpgin

metrics:

irate(node_vmstat_pgpgin{instance=~"$node:$port",job=~"$job"}[5m])

Pagesout - 从启动到现在换出的内存页数的速率(5分钟内) /proc/vmstat pgpgout

metrics:

irate(node_vmstat_pgpgout{instance=~"$node:$port",job=~"$job"}[5m])

2. Memory Pages Swap In / Out

type: Graph
Unit: short
Label: Pages
Pswpin - 数据从磁盘交换区装入内存的速率(5分钟内) /proc/vmstat pswpin

metrics:

irate(node_vmstat_pswpin{instance=~"$node:$port",job=~"$job"}[5m])

Pswpout - 数据从内存转储到磁盘交换区的速率(5分钟内) /proc/vmstat pswpout

metrics:

irate(node_vmstat_pswpout{instance=~"$node:$port",job=~"$job"}[5m])

3. Memory Page Operations

type: Graph
Unit: short
Label: Pages
Pgdeactivate - 激活的平均页数(5分钟内) /proc/vmstat pgdeactivate

metrics:

irate(node_vmstat_pgdeactivate{instance=~"$node:$port",job=~"$job"}[5m])

Pgfree - 释放的平均页数(5分钟内) /proc/vmstat pgfree

metrics:

irate(node_vmstat_pgfree{instance=~"$node:$port",job=~"$job"}[5m])

Pgactivate - 未激活的平均页数(5分钟内) /proc/vmstat pgactivate

metrics:

irate(node_vmstat_pgactivate{instance=~"$node:$port",job=~"$job"}[5m])

4. Memory Page Faults

type: Graph
Unit: short
Label: Faults
Pgfault - 一级页面和二级页面的平均错误数(5分钟内) /proc/vmstat pgfault

metrics:

irate(node_vmstat_pgfault{instance=~"$node:$port",job=~"$job"}[5m])

Pgmajfault - 一级页面的平均错误数(5分钟内) /proc/vmstat pgmajfault

metrics:

irate(node_vmstat_pgmajfault{instance=~"$node:$port",job=~"$job"}[5m])

Pgminfault - 二级页面的平均错误数(5分钟内)

metrics:

irate(node_vmstat_pgfault{instance=~"$node:$port",job=~"$job"}[5m]) - irate(node_vmstat_pgmajfault{instance=~"$node:$port",job=~"$job"}[5m])

5. Memory Pages Reclaimed

type: Graph
Unit: short
Label: Pages
Kswapd_inodesteal - 由 kswapd 回收用于其它目的的平均页面数(5分钟内) /proc/vmstat kswapd_inodesteal

metrics:

irate(node_vmstat_kswapd_inodesteal{instance=~"$node:$port",job=~"$job"}[5m])

Pgindesteal - 由 inode 释放回收的平均页面数(5分钟内) /proc/vmstat pgindesteal

metrics:

irate(node_vmstat_pginodesteal{instance=~"$node:$port",job=~"$job"}[5m])

6. Memory Calls Reclaimed

type: Graph
Unit: short
Label: Cells
Pageoutrun - 由 kswapd调用来回收的平均页面数(5分钟内) /proc/vmstatpageoutrun

metrics:

irate(node_vmstat_pageoutrun{instance=~"$node:$port",job=~"$job"}[5m])

Allocstall - 请求直接回收的平均页面数(5分钟内) /proc/vmstat allocstall

metrics:

irate(node_vmstat_allocstall{instance=~"$node:$port",job=~"$job"}[5m])

Zone_reclaim_failed - 内存域回收失败的平均页面数(5分钟内) /proc/vmstat zone_reclaim_failed

metrics:

irate(node_vmstat_zone_reclaim_failed{instance=~"$node:$port",job=~"$job"}[5m])

7. Memory Page Rotate /proc/vmstat pgrotated

type: Graph
Unit: short
Label: Pages
Pgrotated - 轮换的平均页面数(5分钟内)

metrics:

irate(node_vmstat_pgrotated{instance=~"$node:$port",job=~"$job"}[5m])

8. Memory Page Drop

type: Graph
Unit: short
Label: Cells
Drop_pagecache - 调用释放缓存的平均页面数(5分钟内) /proc/vmstat drop_pagecache

metrics:

node_vmstat_drop_pagecache{instance=~"$node:$port",job=~"$job"}

Drop_slab - 调用释放 slab 缓存的平均页面数(5分钟内) /proc/vmstat drop_slab

metrics:

node_vmstat_drop_slab{instance=~"$node:$port",job=~"$job"}

9. Memory Scan Slab /proc/vmstat slabs_scanned

type: Graph
Unit: short
Slabs_scanned - 被扫描的 Slab 页面的平均页面数(5分钟内)

metrics:

irate(node_vmstat_slabs_scanned{instance=~"$node:$port",job=~"$job"}[5m])

10. Memory Unevictable Pages

type: Graph
Unit: short
Label: Pages
Unevictable_pgs_cleared - Unevictable pages cleared

metrics:

irate(node_vmstat_unevictable_pgs_cleared{instance=~"$node:$port",job=~"$job"}[5m])

Unevictable_pgs_culled - Unevictable pages culled

metrics:

irate(node_vmstat_unevictable_pgs_culled{instance=~"$node:$port",job=~"$job"}[5m])

Unevictable_pgs_mlocked - Unevictable pages mlocked

metrics:

irate(node_vmstat_unevictable_pgs_mlocked{instance=~"$node:$port",job=~"$job"}[5m])

Unevictable_pgs_munlocked - Unevictable pages munlocked

metrics:

irate(node_vmstat_unevictable_pgs_munlocked{instance=~"$node:$port",job=~"$job"}[5m])

Unevictable_pgs_rescued- Unevictable pages rescued

metrics:

irate(node_vmstat_unevictable_pgs_rescued{instance=~"$node:$port",job=~"$job"}[5m])

Unevictable_pgs_scanned - Unevictable pages scanned

metrics:

irate(node_vmstat_unevictable_pgs_scanned{instance=~"$node:$port",job=~"$job"}[5m])

unevictable_pgs_stranded - Unevictable pages stranded

metrics:

irate(node_vmstat_unevictable_pgs_stranded{instance=~"$node:$port",job=~"$job"}[5m])

11. Memory Page Allocation

type: Graph
Unit: short
Label: Pages
Pgalloc_dma - DMA 存储区分配的平均页数(5分钟内) /proc/vmstat pgalloc_dma

metrics:

irate(node_vmstat_pgalloc_dma{instance=~"$node:$port",job=~"$job"}[5m])

Pgalloc_dma32 - DMA32 存储区分配的平均页数(5分钟内) /proc/vmstat pgalloc_dma32

metrics:

irate(node_vmstat_pgalloc_dma32{instance=~"$node:$port",job=~"$job"}[5m])

Pgalloc_movable - movable 存储区分配的平均页数(5分钟内) /proc/vmstat pgalloc_movable

metrics:

irate(node_vmstat_pgalloc_movable{instance=~"$node:$port",job=~"$job"}[5m])

Pgalloc_normal - 普通存储区分配的平均页数(5分钟内) /proc/vmstat pgalloc_normal

metrics:

irate(node_vmstat_pgalloc_normal{instance=~"$node:$port",job=~"$job"}[5m])

12. Memory Page Refill

type: Graph
Unit: short
Label: Pages
Pgrefill_dma - DMA 再填充的平均页数(5分钟内) /proc/vmstat pgrefill_dma

metrics:

irate(node_vmstat_pgrefill_dma{instance=~"$node:$port",job=~"$job"}[5m])

Pgrefill_dma32 - DMA32 存储区再填充的平均页数(5分钟内) /proc/vmstat pgrefill_dma32

metrics:

irate(node_vmstat_pgrefill_dma32{instance=~"$node:$port",job=~"$job"}[5m])

Pgrefill_movable - movable 存储区再填充的平均页数(5分钟内) /proc/vmstat pgrefill_movable

metrics:

irate(node_vmstat_pgrefill_movable{instance=~"$node:$port",job=~"$job"}[5m])

Pgrefill_normal - 普通存储区再填充的平均页数(5分钟内) /proc/vmstat pgrefill_normal

metrics:

irate(node_vmstat_pgrefill_normal{instance=~"$node:$port",job=~"$job"}[5m])

13. Memory Page Steal Direct

type: Graph
Unit: short
Label: Pages
Pgsteal_direct_dma - DMA 存储区被直接回收用于其它目的的平均页面数(5分钟内) /proc/vmstat pgsteal_direct_dma

metrics:

irate(node_vmstat_pgsteal_direct_dma{instance=~"$node:$port",job=~"$job"}[5m])

Pgsteal_direct_dma32 - DMA32 存储区被直接回收用于其它目的的平均页面数(5分钟内) /proc/vmstat pgsteal_direct_dma32

metrics:

irate(node_vmstat_pgsteal_direct_dma32{instance=~"$node:$port",job=~"$job"}[5m])

Pgsteal_direct_movable - movable 存储区被直接回收用于其它目的的平均页面数(5分钟内) /proc/vmstat pgsteal_direct_movable

metrics:

irate(node_vmstat_pgsteal_direct_movable{instance=~"$node:$port",job=~"$job"}[5m])

Pgsteal_direct_normal - 普通存储区被直接回收用于其它目的的平均页面数(5分钟内) /proc/vmstat pgsteal_direct_normal

metrics:

irate(node_vmstat_pgsteal_direct_normal{instance=~"$node:$port",job=~"$job"}[5m])

14. Memory Page Steal Kswapd

type: Graph
Unit: short
Label: Pages
Pgsteal_kswapd_dma - kswapd 后台进程回收 DMA 存储区用于其它目的的平均页面数(5分钟内) /proc/vmstat pgsteal_kswapd_dma

metrics:

irate(node_vmstat_pgsteal_kswapd_dma{instance=~"$node:$port",job=~"$job"}[5m])

Pgsteal_kswapd_dma32 - kswapd 后台进程回收 DMA32 存储区用于其它目的的平均页面数(5分钟内) /proc/vmstat pgsteal_kswapd_dma32

metrics:

irate(node_vmstat_pgsteal_kswapd_dma32{instance=~"$node:$port",job=~"$job"}[5m])

Pgsteal_kswapd_movable - kswapd 后台进程回收 movable 存储区用于其它目的的平均页面数(5分钟内 /proc/vmstat pgsteal_kswapd_movable

metrics:

irate(node_vmstat_pgsteal_kswapd_movable{instance=~"$node:$port",job=~"$job"}[5m])

Pgsteal_kswapd_normal - swapd后台进程回收普通存储区用于其它目的的平均页面数(5分钟内 /proc/vmstat pgsteal_kswapd_normal

metrics:

irate(node_vmstat_pgsteal_kswapd_normal{instance=~"$node:$port",job=~"$job"}[5m])

15. Memory Scan Direct

type: Graph
Unit: short
Label: Pages
Pgscan_direct_dma - DMA 存储区被直接回收的平均页面数(5分钟内) /proc/vmstat pgscan_direct_dma

metrics:

irate(node_vmstat_pgscan_direct_dma{instance=~"$node:$port",job=~"$job"}[5m])

Pgscan_direct_dma32 - DMA32 存储区被直接回收的平均页面数(5分钟内) /proc/vmstat pgscan_direct_dma32

metrics:

irate(node_vmstat_pgscan_direct_dma32{instance=~"$node:$port",job=~"$job"}[5m])

Pgscan_direct_movable - movable 存储区被直接回收的平均页面数(5分钟内) /proc/vmstat pgscan_direct_movable

metrics:

irate(node_vmstat_pgscan_direct_movable{instance=~"$node:$port",job=~"$job"}[5m])

Pgscan_direct_normal - 普通存储区被直接回收的平均页面数(5分钟内) /proc/vmstat pgscan_direct_normal

metrics:

irate(node_vmstat_pgscan_direct_normal{instance=~"$node:$port",job=~"$job"}[5m])

Pgscan_direct_throttle - throttle 存储区被直接回收的平均页面数(5分钟内) /proc/vmstat pgscan_direct_throttle

metrics:

irate(node_vmstat_pgscan_direct_throttle{instance=~"$node:$port",job=~"$job"}[5m])

16. Memory Scan Kswapd

type: Graph
Unit: short
Label: Pages
Pgscan_kswapd_dma - kswapd 后台进程扫描的 DMA 存储区平均页面数(5分钟内) /proc/vmstat pgscan_kswapd_dma

metrics:

irate(node_vmstat_pgscan_kswapd_dma{instance=~"$node:$port",job=~"$job"}[5m])

Pgscan_kswapd_dma32 - kswapd 后台进程扫描的 DMA32 存储区平均页面数(5分钟内) /proc/vmstat pgscan_kswapd_dma32

metrics:

irate(node_vmstat_pgscan_kswapd_dma32{instance=~"$node:$port",job=~"$job"}[5m])

Pgscan_kswapd_movable - kswapd 后台进程扫描的 movable 存储区平均页面数(5分钟内) /proc/vmstat pgscan_kswapd_movable

metrics:

irate(node_vmstat_pgscan_kswapd_movable{instance=~"$node:$port",job=~"$job"}[5m])

Pgscan_kswapd_normal - kswapd 后台进程扫描的普通存储区平均页面数(5分钟内） /proc/vmstat pgscan_kswapd_normal

metrics:

irate(node_vmstat_pgscan_kswapd_normal{instance=~"$node:$port",job=~"$job"}[5m])

17. Memory Page Compact

type: Graph
Unit: short
Label: Pages
Compact_free_scanned - 扫描由压缩守护程序释放的页面 /proc/vmstat compact_free_scanned

metrics:

irate(node_vmstat_compact_free_scanned{instance=~"$node:$port",job=~"$job"}[5m])

Compact_isolated - 用于内存压缩隔离的页面 /proc/vmstat compact_isolated

metrics:

irate(node_vmstat_compact_isolated{instance=~"$node:$port",job=~"$job"}[5m])

Compact_migrate_scanned - 通过内存压缩守护程序扫描以进行迁移的页面 /proc/vmstat compact_migrate_scanned

metrics:

irate(node_vmstat_compact_migrate_scanned{instance=~"$node:$port",job=~"$job"}[5m])

18. Memory Compactions 内存紧缩

type: Graph
Unit: short
Label: Compactions
Compact_fail - 高阶分配的内存碎片整理失败的页面数(5分钟内) /proc/vmstat compact_fail

metrics:

irate(node_vmstat_compact_fail{instance=~"$node:$port",job=~"$job"}[5m])

Compact_stall - 开始执行内存碎片失败的页面数(5分钟内) /proc/vmstat compact_stall

metrics:

irate(node_vmstat_compact_stall{instance=~"$node:$port",job=~"$job"}[5m])

Compact_sucess - 高阶分配的内存碎片整理成功的页面数(5分钟内)

metrics:

irate(node_vmstat_compact_success{instance=~"$node:$port",job=~"$job"}[5m])

19. Memory Kswapd Watermark

type: Graph
Unit: short
Label: Counter
Kswapd_high_wmark_hit_quickly - 剩余内存达到 high 的水位线的时间 /proc/vmstat kswapd_high_wmark_hit_quickly

metrics:

node_vmstat_kswapd_high_wmark_hit_quickly{instance=~"$node:$port",job=~"$job"}

Kswapd_low_wmark_hit_quickly - - 剩余内存达到 low 的水位线的时间 /proc/vmstat kswapd_low_wmark_hit_quickly

metrics:

node_vmstat_kswapd_low_wmark_hit_quickly{instance=~"$node:$port",job=~"$job"}

20. Memory Buddy Alloc

type: Graph
Unit: short
Label: Allocations
Htlb_buddy_alloc_fail - buddy 给 hugetlb 分配失败的次数 /proc/vmstat htlb_buddy_alloc_fail

metrics:

node_vmstat_htlb_buddy_alloc_fail{instance=~"$node:$port",job=~"$job"}

Htlb_buddy_alloc_success - buddy 给 hugetlb 分配成功的次数 /proc/vmstat htlb_buddy_alloc_success

metrics:

node_vmstat_htlb_buddy_alloc_success{instance=~"$node:$port",job=~"$job"}

21. Memory Numa Allocations

type: Graph
Unit: short
Label: Allocations
Numa_foreign - 计划使用其他节点内存但是却使用本地内存次数 /proc/vmstat numa_foreign

metrics:

irate(node_vmstat_numa_foreign{instance=~"$node:$port",job=~"$job"}[5m])

Numa_hit - 使用本节点内存次数 /proc/vmstat numa_hit

metrics:

irate(node_vmstat_numa_hit{instance=~"$node:$port",job=~"$job"}[5m])

Numa_interleave - 交叉分配使用的内存中使用本节点的内存次数 /proc/vmstat numa_interleave

metrics:

irate(node_vmstat_numa_interleave{instance=~"$node:$port",job=~"$job"}[5m])

Numa_local - 在本节点运行的程序使用本节点内存次数 /proc/vmstat numa_local

metrics:

irate(node_vmstat_numa_local{instance=~"$node:$port",job=~"$job"}[5m])

Numa_miss - 计划使用本节点内存而被调度到其他节点次数 /proc/vmstat numa_miss

metrics:

irate(node_vmstat_numa_miss{instance=~"$node:$port",job=~"$job"}[5m])

Numa_other - 在其他节点运行的程序使用本节点内存次数 /proc/vmstat numa_other

metrics:

irate(node_vmstat_numa_other{instance=~"$node:$port",job=~"$job"}[5m])

22. Memory Numa Page Migrations

type: Graph
Unit: short
Label: Pages
Numa_pages_migrated - NUMA page 数 /proc/vmstat numa_pages_migrated

metrics:

irate(node_vmstat_numa_pages_migrated{instance=~"$node:$port",job=~"$job"}[5m])

Pgmigrate_fail - 迁移失败的页面数 /proc/vmstat pgmigrate_fail

metrics:

irate(node_vmstat_pgmigrate_fail{instance=~"$node:$port",job=~"$job"}[5m])

Pgmigrate_success - 成功迁移的页面数 /proc/vmstat pgmigrate_success

metrics:

irate(node_vmstat_pgmigrate_success{instance=~"$node:$port",job=~"$job"}[5m])

23. Memory Numa Hints

type: Graph
Unit: short
Label: Hints
Numa_hint_faults - NUMA hint faults trapped

metrics:

irate(node_vmstat_numa_hint_faults{instance=~"$node:$port",job=~"$job"}[5m])

Numa_hint_faults_local - Hinting faults to local nodes

metrics:

irate(node_vmstat_numa_hint_faults_local{instance=~"$node:$port",job=~"$job"}[5m])

24. Memory Numa Table Updates

type: Graph
Unit: short
Label: Updates
Numa_pte_updates - NUMA page table entry updates

metrics:

irate(node_vmstat_numa_pte_updates{instance=~"$node:$port",job=~"$job"}[5m])

Numa_huge_pte_updates - NUMA huge page table entry updates

metrics:

irate(node_vmstat_numa_huge_pte_updates{instance=~"$node:$port",job=~"$job"}[5m])

25. Memory THP Splits

type: Graph
Unit: short
Label: Splits
Thp_split - 大型页面分割成多个常规页面 /proc/vmstat thp_split

metrics:

irate(node_vmstat_thp_split{instance=~"$node:$port",job=~"$job"}[5m])

26. Memory Workingset

type: Graph
Unit: short
Label: Counter
Workingset_activate - Page activations to form the working set

metrics:

irate(node_vmstat_workingset_activate{instance=~"$node:$port",job=~"$job"}[5m])

Workingset_nodereclaim - NUMA node working set page reclaims

metrics:

irate(node_vmstat_workingset_nodereclaim{instance=~"$node:$port",job=~"$job"}[5m])

Workingset_refault - Refaults of previously evicted pages

metrics:

irate(node_vmstat_workingset_refault{instance=~"$node:$port",job=~"$job"}[5m])

27. Memory THP Allocations

type: Graph
Unit: short
Label: Allocations
Thp_collapse_alloc - Transparent huge page collapse allocations

metrics:

irate(node_vmstat_thp_collapse_alloc{instance=~"$node:$port",job=~"$job"}[5m])

Thp_collapse_alloc_failed - Transparent huge page collapse allocation failures

metrics:

irate(node_vmstat_thp_collapse_alloc_failed{instance=~"$node:$port",job=~"$job"}[5m])

Thp_zero_page_alloc - Transparent huge page zeroed page allocations

metrics:

irate(node_vmstat_thp_zero_page_alloc{instance=~"$node:$port",job=~"$job"}[5m])

Thp_zero_page_alloc_failed - Transparent huge page zeroed page allocation failures

metrics:

irate(node_vmstat_thp_zero_page_alloc_failed{instance=~"$node:$port",job=~"$job"}[5m])

Thp_fault_alloc - Transparent huge page fault allocations

metrics:

irate(node_vmstat_thp_fault_alloc{instance=~"$node:$port",job=~"$job"}[5m])

Thp_fault_fallback - Transparent huge page fault fallbacks

metrics:

irate(node_vmstat_thp_fault_fallback{instance=~"$node:$port",job=~"$job"}[5m])