zabbix优化之partition

最新推荐文章于 2022-04-24 17:26:39 发布

转载最新推荐文章于 2022-04-24 17:26:39 发布 · 325 阅读

0 ·

CC 4.0 BY-SA版权

原文链接：https://my.oschina.net/u/3023401/blog/840724

文章标签：

#运维 #数据库 #python

2019独角兽企业重金招聘Python工程师标准>>>

一、对zabbix进行partition

1.1 我的思路

1、在不影响zabbix业务正常写入的情况下，在线搭建一个从库

2、对从库历史数据进行表分区，查看每个分区业务增长量

3、根据业务增长量和字段大小规划数据库容量，然后进行合理的分区

4、从库运行一段时间，查看从库状态，状态正常就切换zabbix连接从库

1.2 对历史数据进行分析

线上情况如上图所示，线上items按2W个计算

1.2.1 历史数据的保存

zabbix的数据库大小取决于NVPS（每秒处理的平均数量），我们线上按2W个监控项，刷新周期按60s，那么每秒处理的数据为20000/60=333(条)，说明此刻每秒将会向数据库新增加333条数据，可以设置历史数据库保存天数，

计算公式：历史数据大小=天数*每秒处理的数据*24h*3600s*50Bytes

1.2.2 趋势数据的保存

每个Items趋势的数据大约90B,大小取决于数据库类型。trends每小时收集一次数据，

计算公式为：趋势数据大小=监控项*24h*90B

1.3 在从库进行操作

3.4.1 index change

Zabbix3.2不用进行这步操作，可惜我是3.0

如果数据量特别大，这可能花费很长时间，几秒，几分钟，几个小时，几天都是可能的

多mysql版本中，DDL不是online的，5.6之前

需要修改的表结构如下

mysql> Alter table trends drop primary key, add index (itemid),add index (clock);

mysql> Alter table trends_uint drop primary key, add index (itemid),add index (clock);

3.4.2 Stored Procedures

DELIMITER $$

CREATE PROCEDURE `partition_create`(SCHEMANAME varchar(64), TABLENAME varchar(64), PARTITIONNAME varchar(64), CLOCK int)

BEGIN

        /*

           SCHEMANAME = The DB schema in which to make changes

           TABLENAME = The table with partitions to potentially delete

           PARTITIONNAME = The name of the partition to create

        */

        /*

           Verify that the partition does not already exist

        */



        DECLARE RETROWS INT;

        SELECT COUNT(1) INTO RETROWS

        FROM information_schema.partitions

        WHERE table_schema = SCHEMANAME AND table_name = TABLENAME AND partition_description >= CLOCK;



        IF RETROWS = 0 THEN

                /*

                   1. Print a message indicating that a partition was created.

                   2. Create the SQL to create the partition.

                   3. Execute the SQL from #2.

                */

                SELECT CONCAT( "partition_create(", SCHEMANAME, ",", TABLENAME, ",", PARTITIONNAME, ",", CLOCK, ")" ) AS msg;

                SET @sql = CONCAT( 'ALTER TABLE ', SCHEMANAME, '.', TABLENAME, ' ADD PARTITION (PARTITION ', PARTITIONNAME, ' VALUES LESS THAN (', CLOCK, '));' );

                PREPARE STMT FROM @sql;

                EXECUTE STMT;

                DEALLOCATE PREPARE STMT;

        END IF;

END$$

DELIMITER ;

DELIMITER $$

CREATE PROCEDURE `partition_drop`(SCHEMANAME VARCHAR(64), TABLENAME VARCHAR(64), DELETE_BELOW_PARTITION_DATE BIGINT)

BEGIN

        /*

           SCHEMANAME = The DB schema in which to make changes

           TABLENAME = The table with partitions to potentially delete

           DELETE_BELOW_PARTITION_DATE = Delete any partitions with names that are dates older than this one (yyyy-mm-dd)

        */

        DECLARE done INT DEFAULT FALSE;

        DECLARE drop_part_name VARCHAR(16);



        /*

           Get a list of all the partitions that are older than the date

           in DELETE_BELOW_PARTITION_DATE.  All partitions are prefixed with

           a "p", so use SUBSTRING TO get rid of that character.

        */

        DECLARE myCursor CURSOR FOR

                SELECT partition_name

                FROM information_schema.partitions

                WHERE table_schema = SCHEMANAME AND table_name = TABLENAME AND CAST(SUBSTRING(partition_name FROM 2) AS UNSIGNED) < DELETE_BELOW_PARTITION_DATE;

        DECLARE CONTINUE HANDLER FOR NOT FOUND SET done = TRUE;



        /*

           Create the basics for when we need to drop the partition.  Also, create

           @drop_partitions to hold a comma-delimited list of all partitions that

           should be deleted.

        */

        SET @alter_header = CONCAT("ALTER TABLE ", SCHEMANAME, ".", TABLENAME, " DROP PARTITION ");

        SET @drop_partitions = "";



        /*

           Start looping through all the partitions that are too old.

        */

        OPEN myCursor;

        read_loop: LOOP

                FETCH myCursor INTO drop_part_name;

                IF done THEN

                        LEAVE read_loop;

                END IF;

                SET @drop_partitions = IF(@drop_partitions = "", drop_part_name, CONCAT(@drop_partitions, ",", drop_part_name));

        END LOOP;

        IF @drop_partitions != "" THEN

                /*

                   1. Build the SQL to drop all the necessary partitions.

                   2. Run the SQL to drop the partitions.

                   3. Print out the table partitions that were deleted.

                */

                SET @full_sql = CONCAT(@alter_header, @drop_partitions, ";");

                PREPARE STMT FROM @full_sql;

                EXECUTE STMT;

                DEALLOCATE PREPARE STMT;



                SELECT CONCAT(SCHEMANAME, ".", TABLENAME) AS `table`, @drop_partitions AS `partitions_deleted`;

        ELSE

                /*

                   No partitions are being deleted, so print out "N/A" (Not applicable) to indicate

                   that no changes were made.

                */

                SELECT CONCAT(SCHEMANAME, ".", TABLENAME) AS `table`, "N/A" AS `partitions_deleted`;

        END IF;

END$$

DELIMITER ;

DELIMITER $$

CREATE PROCEDURE `partition_maintenance`(SCHEMA_NAME VARCHAR(32), TABLE_NAME VARCHAR(32), KEEP_DATA_DAYS INT, HOURLY_INTERVAL INT, CREATE_NEXT_INTERVALS INT)

BEGIN

        DECLARE OLDER_THAN_PARTITION_DATE VARCHAR(16);

        DECLARE PARTITION_NAME VARCHAR(16);

        DECLARE OLD_PARTITION_NAME VARCHAR(16);

        DECLARE LESS_THAN_TIMESTAMP INT;

        DECLARE CUR_TIME INT;



        CALL partition_verify(SCHEMA_NAME, TABLE_NAME, HOURLY_INTERVAL);

        SET CUR_TIME = UNIX_TIMESTAMP(DATE_FORMAT(NOW(), '%Y-%m-%d 00:00:00'));



        SET @__interval = 1;

        create_loop: LOOP

                IF @__interval > CREATE_NEXT_INTERVALS THEN

                        LEAVE create_loop;

                END IF;



                SET LESS_THAN_TIMESTAMP = CUR_TIME + (HOURLY_INTERVAL * @__interval * 3600);

                SET PARTITION_NAME = FROM_UNIXTIME(CUR_TIME + HOURLY_INTERVAL * (@__interval - 1) * 3600, 'p%Y%m%d%H00');

                IF(PARTITION_NAME != OLD_PARTITION_NAME) THEN

                        CALL partition_create(SCHEMA_NAME, TABLE_NAME, PARTITION_NAME, LESS_THAN_TIMESTAMP);

                END IF;

                SET @__interval=@__interval+1;

                SET OLD_PARTITION_NAME = PARTITION_NAME;

        END LOOP;



        SET OLDER_THAN_PARTITION_DATE=DATE_FORMAT(DATE_SUB(NOW(), INTERVAL KEEP_DATA_DAYS DAY), '%Y%m%d0000');

        CALL partition_drop(SCHEMA_NAME, TABLE_NAME, OLDER_THAN_PARTITION_DATE);



END$$

DELIMITER ;

DELIMITER $$

CREATE PROCEDURE `partition_verify`(SCHEMANAME VARCHAR(64), TABLENAME VARCHAR(64), HOURLYINTERVAL INT(11))

BEGIN

        DECLARE PARTITION_NAME VARCHAR(16);

        DECLARE RETROWS INT(11);

        DECLARE FUTURE_TIMESTAMP TIMESTAMP;



        /*

         * Check if any partitions exist for the given SCHEMANAME.TABLENAME.

         */

        SELECT COUNT(1) INTO RETROWS

        FROM information_schema.partitions

        WHERE table_schema = SCHEMANAME AND table_name = TABLENAME AND partition_name IS NULL;



        /*

         * If partitions do not exist, go ahead and partition the table

         */

        IF RETROWS = 1 THEN

                /*

                 * Take the current date at 00:00:00 and add HOURLYINTERVAL to it.  This is the timestamp below which we will store values.

                 * We begin partitioning based on the beginning of a day.  This is because we don't want to generate a random partition

                 * that won't necessarily fall in line with the desired partition naming (ie: if the hour interval is 24 hours, we could

                 * end up creating a partition now named "p201403270600" when all other partitions will be like "p201403280000").

                 */

                SET FUTURE_TIMESTAMP = TIMESTAMPADD(HOUR, HOURLYINTERVAL, CONCAT(CURDATE(), " ", '00:00:00'));

                SET PARTITION_NAME = DATE_FORMAT(CURDATE(), 'p%Y%m%d%H00');



                -- Create the partitioning query

                SET @__PARTITION_SQL = CONCAT("ALTER TABLE ", SCHEMANAME, ".", TABLENAME, " PARTITION BY RANGE(`clock`)");

                SET @__PARTITION_SQL = CONCAT(@__PARTITION_SQL, "(PARTITION ", PARTITION_NAME, " VALUES LESS THAN (", UNIX_TIMESTAMP(FUTURE_TIMESTAMP), "));");



                -- Run the partitioning query

                PREPARE STMT FROM @__PARTITION_SQL;

                EXECUTE STMT;

                DEALLOCATE PREPARE STMT;

        END IF;

END$$

DELIMITER ;

3.4.3 Using the stored procedures

你不需要理解这些是怎么工作的，下面是一个例子：

CALL partition_maintenance('<zabbix_db_name>', '<table_name>', <days_to_keep_data>, <hourly_interval>, <num_future_intervals_to_create>)；

下面是一个例子

mysql> CALL partition_maintenance('zabbix', 'history', 28, 24, 14);

+-----------------------------------------------------------+

| msg                                                       |

+-----------------------------------------------------------+

| partition_create(zabbix,history,p201404160000,1397718000) |

+-----------------------------------------------------------+

1 row in set (0.39 sec)

……

…..

3.4.4 Improving procedure calls

DELIMITER $$

CREATE PROCEDURE `partition_maintenance_all`(SCHEMA_NAME VARCHAR(32))

BEGIN

                CALL partition_maintenance(SCHEMA_NAME, 'history', 30, 24, 30);

                CALL partition_maintenance(SCHEMA_NAME, 'history_uint', 30, 24, 30);

                CALL partition_maintenance(SCHEMA_NAME, 'trends', 720, 1440, 30);

                CALL partition_maintenance(SCHEMA_NAME, 'trends_uint', 720, 1440, 30);

END$$

DELIMITER ;

然后统一调用

mysql> CALL partition_maintenance_all('zabbix');

+----------------+--------------------+

| table          | partitions_deleted |

+----------------+--------------------+

| zabbix.history | N/A                |

+----------------+--------------------+

1 row in set (0.01 sec)



....

....

....

3.4.5 Housekeeper changes

使用zabbix2.0.x的用户比较悲惨，自己体会吧

Zabbix2.2以上版本

进入zabbix页面，"Administration" -> "General" section.选择Housekeeping

3.4.5 All done

00 00 29 * * mysql zabbix -e "CALL partition_maintenance_all('zabbix');"

参考文档：

https://www.zabbix.org/wiki/Docs/howto/mysql_partition

为了方便大家交流，本人开通了微信公众号，和QQ群291519319。喜欢技术的一起来交流吧

转载于:https://my.oschina.net/u/3023401/blog/840724