[CONTENT] Linux recommended swap size

Content:

For RHEL 6, RHEL 7, RHEL 8, RHEL 9

RAM sizeRecommended swap sizeRecommended swap size if allowing for hibernation
From 0 to 2GB 2 times the RAM size3 times the RAM size
From 2GB to 8GBThe same size of the RAM2 times the RAM size
From 8GB to 64GBAt least 4GB1.5 times the RAM size
From 64GBAt least 4GBHibernation is not recommended
RHEL 6, RHEL 7, RHEL 8, RHEL 9 Recommended Swap Size Table

Note: A 100GB swap is recommended if system with over 140 logical processes or over 3TB RAM

Reference:

https://access.redhat.com/solutions/15244

[步骤] Linux page_owner 排查工具的使用 (记录内存的使用情况)

正文:

步骤一:查看 page_owner 是否开启

1.1 通过 dmesg 命令查看 page_owner 是否开启

# dmesg | grep page_owner
[    1.149165] page_owner is disabled

(补充:当显示此类信息时则 page_owner 没有开启)

1.2 通过 /sys/kernel/debug/ 目录查看 page_ownerr 是否开启

# ls -l /sys/kernel/debug/page_owner
ls: cannot access /sys/kernel/debug/page_owner: No such file or directory.

(补充:当 /sys/kernel/debug/page_owner 文件不存在时则 page_owner 没有开启)

步骤二:开启 page_owner

2.1 开启 page_owner

# grubby --args="page_owner=on" --update-kernel=0

(注意:开启 page_owner 会额外占用一定量的内存)

2.2 重启系统

# reboot

2.3 确认 page_owner 已经开启

2.3.1 通过 dmesg 命令确认 page_owner 是否开启
# dmesg | grep page_owner
[    0.000000] Command line: BOOT_IMAGE=(hd0,gpt2)/vmlinuz-4.18.0-425.19.2.el8_7.x86_64 root=/dev/mapper/rootvg-rootlv ro ipv6.disable=1 audit=1 audit_backlog_limit=8192 crashkernel=auto resume=/dev/mapper/rootvg-swaplv rd.lvm.lv=rootvg/rootlv rd.lvm.lv=rootvg/swaplv rhgb quiet rd.shell=0 page_owner=on
[    0.000000] Kernel command line: BOOT_IMAGE=(hd0,gpt2)/vmlinuz-4.18.0-425.19.2.el8_7.x86_64 root=/dev/mapper/rootvg-rootlv ro ipv6.disable=1 audit=1 audit_backlog_limit=8192 crashkernel=auto resume=/dev/mapper/rootvg-swaplv rd.lvm.lv=rootvg/rootlv rd.lvm.lv=rootvg/swaplv rhgb quiet rd.shell=0 page_owner=o

(补充:当显示此类信息时则 page_owner 已经开启)

2.3.2 通过 /sys/kernel/debug/ 目录确认 page_ownerr 是否开启
# ls -l /sys/kernel/debug/page_owner
-r--------. 1 root root 0 Apr 13 14:36 /sys/kernel/debug/page_owner

(补充:当 /sys/kernel/debug/page_owner 文件存在时则 page_owner 已经开启)

步骤三:分析 page_owner 产生的记录

3.1 导出 page_owner 产生的记录

# cat /sys/kernel/debug/page_owner > page_owner_full.txt

(补充:这里以将 page_owner 产生的记录导出到名为 page_owner_full.txt 文件为例)


注意:
1) 此命令会产生体积很巨大的文件
2) 此命令会持续执行直到手动取消
3) 可以通过同时按下 “ctrl” 键和 “C” 键或者使用 kill 命令取消
4) 如果内存变化较快则可以让此命令多执行一会,反之则可以少执行一会

3.2 解析 page_owner 产生的记录

# page_owner_sort page_owner_full.txt sorted_page_owner.txt
loaded 42903
sorting ....
culling

(补充:这里以分析名为 page_owner_full.txt 的文件并将分析结果导入到 sorted_page_owner.txt 文件为例)

3.3 查看 page_owner 产生的记录

# less sorted_page_owner.txt
1 times:
Page allocated via order 0, mask 0x0(), pid 1, tgid 1 (swapper/0), ts 48952109 ns, free_ts 0 ns
PFN 4096 type Unmovable Block 8 type Unmovable Flags 0xfffffc0000100(slab|node=0|zone=1|lastcpupid=0x1fffff)
 register_early_stack+0x28/0x60
 init_page_owner+0x30/0x2d0
 kernel_init_freeable+0x13c/0x232
 kernel_init+0xa/0x108

1 times:
Page allocated via order 0, mask 0x0(), pid 1, tgid 1 (swapper/0), ts 48952566 ns, free_ts 0 ns
PFN 4097 type Unmovable Block 8 type Unmovable Flags 0xfffffc0000100(slab|node=0|zone=1|lastcpupid=0x1fffff)
 register_early_stack+0x28/0x60
 init_page_owner+0x30/0x2d0
 kernel_init_freeable+0x13c/0x232
 kernel_init+0xa/0x108
......

(补充:这里以查看名为 sorted_page_owner.txt 文件里的分析结果为例)

步骤四:关闭 page_owner

4.1 开启 page_owner

# grubby --remove-args="page_owner=on" --update-kernel=0

(注意:关闭 page_owner 会额外释放一定量的内存)

4.2 重启系统

# reboot

4.3 确认 page_owner 已关闭

4.3.1 通过 dmesg 命令确认 page_owner 是否关闭
# dmesg | grep page_owner
[    2.022585] page_owner is disabled

(补充:当显示此类信息时则 page_owner 没有开启)

4.3.2 通过 /sys/kernel/debug/ 目录确认 page_ownerr 是否关闭
# ls -l /sys/kernel/debug/page_owner
ls: cannot access '/sys/kernel/debug/page_owner': No such file or directory

(补充:当 /sys/kernel/debug/page_owner 文件不存在时则 page_owner 没有开启)

参考文献:

https://access.redhat.com/solutions/5609521