2007年7月20日 星期五

MySQL偽memory leak問題

Problem

我用的server是Debian 64bit (AMD64), DB server是 5.0.32,有8G memory,但MySQL使用一陣子後,用top和free會看到8G的memory都被吃掉,但沒有任何process用到這麼多memory,而且swap沒有被用到,關掉MySQL daemon後memory的用量也沒改變。

Solution

查到兩篇類似的苦主:

從第二篇裡找到解法:

Despite the fact that “top”, “/proc/meminfor”, “free” show me that 98% of all memory on server is used that memory is not really used by MySQL nor kernel (as I can see from /proc/slabinfo).

I wrote simple C app. which tries to allocate as much memory as possible calling malloc() in a loop I found how much free memory _really_ available. Looks like it is Linux reporting problem, but same time this _only_ happen to me with memory previously used by MySQL server. I can see memory returned back to free memory pool as soon as other applications exit.

上面說是top和free報錯數據,MySQL query結束後,query用掉的memory有還回去,但top和free的統計數據裡卻沒有減少。

我也寫了個 C app來試,如他所言一般,程式結束後memory回來了,但查了半天找不到單一process的maximum memory usage定義在那,有人說在limits.h裡,INT_MAX就是maximum memory,但顯然是錯的,我設到比它大的值,anyway,下面是我「要回」memory的 code,因為設到15*會超出上限被killed,只好設12*並同時跑多個,但想想這個數字也不太對,它比8G大,但跑起來大概只用了2G吧,不知道process到底怎麼跑的

#include <stdio.h> #include <stdlib.h> enum { MAX = 12000000000 }; char s[MAX]; int main(void) { int i; for (i=0; i<MAX; i++) s[i] = i; sleep(10); return 0; }

另外聽小豬說在Debian的MySQL 5.1上沒這問題,也許是已修正的bug吧。

備註:query cache相關指令

http://dev.mysql.com/doc/refman/5.0/en/flush.html
http://dev.mysql.com/doc/refman/5.0/en/query-cache-status-and-maintenance.html

The RESET QUERY CACHE statement removes all query results from the query cache. The FLUSH TABLES statement also does this.

還有可以用SHOW VARIABLES查環境設定,像是myisam_max_sort_file_size、max_binlog_cache_size。

沒有留言:

張貼留言