zabbix 由于历史数据过大, 因此导致磁盘空间暴涨, 下面是结局方法步骤
1. 停止 ZABBIX SERER 操作
[plain] view plaincopyprint?
[root@gd02-qa-plxt2-nodomain-web-95 ~]# killall zabbix_server
[root@gd02-qa-plxt2-nodomain-web-95 ~]# lsof -i:10051
2. 停止 mysql 操作
[plain] view plaincopyprint?
[root@gd02-qa-plxt2-nodomain-web-96 dbdat]# mysqladmin -u root -p -h 127.0.0.1 shutdown
3. 修改 my.cnf
添加 skip-new 参数, 目标可用缩减 innodb 磁盘空间
4. 重启启动 mysql
[plain] view plaincopyprint?
/apps/svr/mysql/bin/mysqld_safe --defaults-file=/apps/conf/mysql5.6/my.cnf --ledir=/apps/svr/mysql/bin --basedir=/apps/svr/mysql/ --datadir=/apps/dbdat/mysql5_data --user=apps &
[root@gd02-qa-plxt2-nodomain-web-96 apps]# lsof -i:3306
COMMAND PID USER FD TYPE DEVICE SIZE/OFF NODE NAME
mysqld 25527 apps 11u IPv4 29371110 0t0 TCP *:mysql (LISTEN)
5. 分析 history 表
[plain] view plaincopyprint?
mysql> desc history;
+--------+---------------------+------+-----+---------+-------+
| Field | Type | Null | Key | Default | Extra |
+--------+---------------------+------+-----+---------+-------+
| itemid | bigint(20) unsigned | NO | MUL | NULL | |
| clock | int(11) | NO | | 0 | |
| value | double(16,4) | NO | | 0.0000 | |
| ns | int(11) | NO | | 0 | |
+--------+---------------------+------+-----+---------+-------+
4 rows in set (0.00 sec)
mysql> select max(itemid) from history;
+-------------+
| max(itemid) |
+-------------+
| 46582 |
+-------------+
1 row in set (0.00 sec)
mysql> select * from history where itemid=46582 limit 1, 20;
+--------+------------+--------+-----------+
| itemid | clock | value | ns |
+--------+------------+--------+-----------+
| 46582 | 1396361332 | 0.0000 | 81875000 |
| 46582 | 1396361362 | 0.0000 | 768297000 |
| 46582 | 1396361392 | 0.0000 | 656787000 |
| 46582 | 1396361422 | 0.0000 | 665169000 |
| 46582 | 1396361452 | 0.0000 | 973570000 |
| 46582 | 1396361482 | 0.0000 | 625619000 |
| 46582 | 1396361512 | 0.0000 | 292743000 |
| 46582 | 1396361543 | 0.0000 | 340000 |
| 46582 | 1396361572 | 0.0000 | 15651000 |
| 46582 | 1396361602 | 0.0000 | 153264000 |
| 46582 | 1396361632 | 0.0000 | 79316000 |
| 46582 | 1396361662 | 0.0000 | 308107000 |
| 46582 | 1396361692 | 0.0000 | 237146000 |
| 46582 | 1396361722 | 0.0000 | 108810000 |
| 46582 | 1396361752 | 0.0000 | 419398000 |
| 46582 | 1396361782 | 0.0000 | 284113000 |
| 46582 | 1396361812 | 0.0000 | 254230000 |
| 46582 | 1396361842 | 0.0000 | 145938000 |
| 46582 | 1396361872 | 0.0000 | 403163000 |
| 46582 | 1396361902 | 0.3300 | 193302000 |
+--------+------------+--------+-----------+
20 rows in set (0.01 sec)
6. 删除两周前数据方法
取得时间戳, 时间只保留至 2014 3 25 日
[plain] view plaincopyprint?
[root@gd02-zabbix-db-research api]# date +%s -d "Mar 25, 2014 00:00:00"
1395676800
删除 history, history_unit 表方法
[plain] view plaincopyprint?
mysql> delete from history where clock < 1395676800;
Query OK, 8961912 rows affected (17 min 22.06 sec)
mysql> delete from history_uint where clock < 1395676800;
Query OK, 7789494 rows affected (21 min 38.02 sec)
尝试对表进行缩减发生故障
[plain] view plaincopyprint?
mysql> optimize table history_uint;
ERROR 1114 (HY000): The table ‘#sql-63b7_2d‘ is full
mysql> quit
Bye
[root@gd02-qa-plxt2-nodomain-web-96 mysql5_data]# df -h
文件系统 容量 已用 可用 已用% 挂载点
/dev/vda1 20G 18G 1.3G 94% /
tmpfs 2.0G 0 2.0G 0% /dev/shm
故障原因, 当前 / 下磁盘空间不够.
临时删除 swapfile
[plain] view plaincopyprint?
[root@gd02-qa-plxt2-nodomain-web-96 /]# swapoff -a
[root@gd02-qa-plxt2-nodomain-web-96 /]# rm -rf swapfile
[root@gd02-qa-plxt2-nodomain-web-96 /]# df -h
文件系统 容量 已用 可用 已用% 挂载点
/dev/vda1 20G 15G 3.7G 81% /
tmpfs 2.0G 0 2.0G 0% /dev/shm
原理说明, mysql 执行 optimize 过程中, 生成了临时表见下面文件, mysql 直接吧 history_unit 表复制成临时表再重新改名, 实现空间缩减.
[plain] view plaincopyprint?
[root@gd02-qa-plxt2-nodomain-web-96 zabbix]# ls *63b7_2e* -lh
-rw-rw---- 1 apps apps 8.5K 04-02 12:04 #sql-63b7_2e.frm
-rw-rw---- 1 apps apps 1.1G 04-02 12:10 #sql-63b7_2e.ibd
再次缩减
[plain] view plaincopyprint?
mysql> optimize table history_uint;
Query OK, 40496963 rows affected (20 min 0.04 sec)
Records: 40496963 Duplicates: 0 Warnings: 0
mysql> optimize table history;
Query OK, 45998084 rows affected (21 min 54.99 sec)
Records: 45998084 Duplicates: 0 Warnings: 0
7. 缩减前后文件大小比较
缩减前
[plain] view plaincopyprint?
[root@gd02-qa-plxt2-nodomain-web-96 mysql5_data]# find -size +50M -exec ls -lh {} \;
-rw-rw---- 1 apps apps 80M 04-02 10:29 ./zabbix/events.ibd
-rw-rw---- 1 apps apps 5.1G 04-02 11:44 ./zabbix/history.ibd
-rw-rw---- 1 apps apps 152M 04-02 10:29 ./zabbix/trends.ibd
-rw-rw---- 1 apps apps 4.3G 04-02 11:47 ./zabbix/history_uint.ibd
-rw-rw---- 1 apps apps 328M 04-02 10:29 ./zabbix/trends_uint.ibd
-rw-rw---- 1 apps apps 1.1G 04-02 11:47 ./ibdata1
缩减后
[plain] view plaincopyprint?
[root@gd02-qa-plxt2-nodomain-web-96 mysql5_data]# find -size +50M -exec ls -lh {} \;
-rw-rw---- 1 apps apps 80M 04-02 10:29 ./zabbix/events.ibd
-rw-rw---- 1 apps apps 3.6G 04-02 13:30 ./zabbix/history.ibd
-rw-rw---- 1 apps apps 152M 04-02 10:29 ./zabbix/trends.ibd
-rw-rw---- 1 apps apps 3.2G 04-02 12:24 ./zabbix/history_uint.ibd
-rw-rw---- 1 apps apps 328M 04-02 10:29 ./zabbix/trends_uint.ibd
-rw-rw---- 1 apps apps 1.1G 04-02 13:30 ./ibdata1
8. 重启启动 zabbix, php, nginx, mysql
新问题出现:
当前 zabbix 进行初始化, 会对 mysql 进行大量数据 r/w 操作
因此可能会发生下面警报, 经过 5 分钟后初始化, 下面报警会自动消除, 不用担心.
[plain] view plaincopyprint?
Disk I/O is overloaded on gd02-qa-plxt2-nodomain-web-96.vclound.com
Zabbix history syncer processes more than 75% busy
Zabbix timer processes more than 75% busy
本文出自 “XFICC” 博客，请务必保留此出处http://xficc.blog.51cto.com/1189288/1571282
原文地址：http://xficc.blog.51cto.com/1189288/1571282
