Linux系统OOM killer机制详解

介绍：

Linux下面有个特性叫OOM killer（Out Of Memory killer），会在系统内存耗尽的情况下出现，选择性的干掉一些进程以求释放一些内存。广大从事Linux方面的IT农民工兄弟或多或少都有遇到过吧。

现象：

主机SSH突然无法连接，主机重启要过很久，重启后主机正常，登录查看日志信息，会有大量的 " Out of Memory: Killed process xxx 、Out of Memory: Killed process xxx "

案例：

kernel: lowmem_reserve[]: 0 0 0 0

kernel: Normal free:0kB min:0kB low:0kB high:0kB active:0kB inactive:0kB present:0kB pages_scanned:0 all_unreclaimable? no

kernel: lowmem_reserve[]: 0 0 0 0

kernel: HighMem free:0kB min:128kB low:128kB high:128kB active:0kB inactive:0kB present:0kB pages_scanned:0 all_unreclaimable? no

kernel: lowmem_reserve[]: 0 0 0 0

kernel: DMA: 0*4kB 1*8kB 1*16kB 1*32kB 1*64kB 1*128kB 1*256kB 1*512kB 1*1024kB 0*2048kB 0*4096kB = 2040kB

kernel: DMA32: 22*4kB 4*8kB 1*16kB 1*32kB 0*64kB 0*128kB 1*256kB 1*512kB 1*1024kB 1*2048kB 0*4096kB = 4008kB

kernel: Normal: empty

kernel: HighMem: empty

kernel: 8917 pagecache pages

kernel: Swap cache: add 96509, delete 88464, find 14295/17401, race 0+4

kernel: Free swap = 0kB

kernel: Total swap = 262136kB

kernel: Free swap: 0kB

kernel: 262144 pages of RAM

kernel: 8801 reserved pages

kernel: 45869 pages shared

kernel: 8045 pages swap cached

kernel: Out of memory: Killed process 330, UID 500, (mysqld).

kernel: httpd invoked oom-killer: gfp_mask=0x201d2, order=0, oomkilladj=0

kernel:

kernel: Call Trace:

kernel: [<ffffffff802c2be6>] out_of_memory+0x8b/0x203

kernel:[<ffffffff8020fab8>] __alloc_pages+0x27f/0x308

kernel:[<ffffffff80213a38>] __do_page_cache_readahead+0xc8/0x1af

kernel:[<ffffffff802142a9>] filemap_nopage+0x14c/0x360

kernel:[<ffffffff80208e9d>] __handle_mm_fault+0x444/0x144f

kernel:[<ffffffff80234691>] lock_sock+0xa7/0xb2

kernel:[<ffffffff80232068>] release_sock+0x13/0xbe

kernel:[<ffffffff8022c1bd>] local_bh_enable+0x9/0x9c

kernel:[<ffffffff80234691>] lock_sock+0xa7/0xb2

kernel:[<ffffffff80267d94>] do_page_fault+0xf72/0x131b

kernel: [<ffffffff8041f85e>] move_addr_to_user+0x5d/0x78

kernel: [<ffffffff8041ff45>] sys_accept+0x18c/0x1d2

kernel: [<ffffffff8026082b>] error_exit+0x0/0x6e

kernel:

kernel: Mem-info:

kernel: DMA per-cpu:

kernel: cpu 0 hot: high 0, batch 1 used:0

kernel: cpu 0 cold: high 0, batch 1 used:0

kernel: cpu 1 hot: high 0, batch 1 used:0

kernel: cpu 1 cold: high 0, batch 1 used:0

kernel: DMA32 per-cpu:

kernel: cpu 0 hot: high 186, batch 31 used:46

kernel: cpu 0 cold: high 62, batch 15 used:12

kernel: cpu 1 hot: high 186, batch 31 used:25

kernel: cpu 1 cold: high 62, batch 15 used:14

kernel: Normal per-cpu: empty

kernel: HighMem per-cpu: empty

kernel: Free pages: 6040kB (0kB HighMem)

kernel: Active:120445 inactive:113583 dirty:0 writeback:0 unstable:0 free:1510 slab:4042 mapped-file:548 mapped-anon:234724 pagetables:7852

kernel: DMA free:2040kB min:32kB low:40kB high:48kB active:0kB inactive:0kB present:9060kB pages_scanned:0 all_unreclaimable? yes

kernel: lowmem_reserve[]: 0 994 994 994

kernel: DMA32 free:4000kB min:4016kB low:5020kB high:6024kB active:481780kB inactive:454332kB present:1018080kB pages_scanned:1718135 all_unreclaimable? yes

以上的日志信息标红处就是出现 OOM Killer（Out Of Memory killer）。

原理：

linux oom-killer是一种自我保护机制，当系统分配不出内存时(触发条件)会触发这个机制，由操作系统在己有进程中挑选一个占用内存较多，回收内存收益最大的进程kill掉来释放内存。

系统为每个进程做评估(/proc/<pid>/oom_score中数值最大的进程被kill掉)。

详见oom_killer.c

/**

* badness – calculate a numeric value for how bad this task has been

* @p: task struct of which task we should calculate

* @uptime: current uptime in seconds

*

* The formula used is relatively simple and documented inline in the

* function. The main rationale is that we want to select a good task

* to kill when we run out of memory.

*

* Good in this context means that:

* 1) we lose the minimum amount of work done

* 2) we recover a large amount of memory

* 3) we don’t kill anything innocent of eating tons of memory

* 4) we want to kill the minimum amount of processes (one)

* 5) we try to kill the process the user expects us to kill, this

* algorithm has been meticulously tuned to meet the principle

* of least surprise … (be careful when you change it)

*/

Linux下有3种Overcommit的策略（参考内核文档：vm/overcommit-accounting），可以在/proc/sys/vm/overcommit_memory配置。取0,1和2三个值，默认是0。

0：启发式策略，比较严重的Overcommit将不能得逞，比如你突然申请了128TB的内存。而轻微的Overcommit将被允许。另外，root能Overcommit的值比普通用户要稍微多些。

1：永远允许Overcommit，这种策略适合那些不能承受内存分配失败的应用，比如某些科学计算应用。

2：永远禁止Overcommit，在这个情况下，系统所能分配的内存不会超过swap+RAM*系数（/proc/sys/vm/overcmmit_ratio，默认50%，你可以调整），如果这么多资源已经用光，那么后面任何尝试申请内存的行为都会返回错误，这通常意味着此时没法运行任何新程序。