Reputation: 23
CREATE TABLE `tvnotif` (
`id` int(11) NOT NULL AUTO_INCREMENT,
`pingId` int(11) DEFAULT NULL,
`token` varchar(45) COLLATE utf8_bin DEFAULT NULL,
`summary` varchar(45) COLLATE utf8_bin DEFAULT NULL,
`startTime` int(11) DEFAULT NULL,
`endTime` int(11) DEFAULT NULL,
`processed` int(1) DEFAULT '0',
`created` timestamp NULL DEFAULT CURRENT_TIMESTAMP,
`modified` timestamp NULL DEFAULT CURRENT_TIMESTAMP ON UPDATE CURRENT_TIMESTAMP,
PRIMARY KEY (`id`),
KEY `processedIndex` (`processed`),
KEY `summaryIndex` (`summary`),
KEY `tokenIndex` (`token`)
) ENGINE=InnoDB AUTO_INCREMENT=18297898 DEFAULT CHARSET=utf8 COLLATE=utf8_bin;
CREATE TABLE `vv_us` (
`id` int(11) NOT NULL AUTO_INCREMENT,
`athleteid` int(11) DEFAULT NULL,
`token` varchar(45) COLLATE utf8_bin DEFAULT NULL,
`secret` varchar(45) COLLATE utf8_bin DEFAULT NULL,
`active` int(1) DEFAULT '1',
`created` timestamp NULL DEFAULT CURRENT_TIMESTAMP,
`modified` timestamp NULL DEFAULT CURRENT_TIMESTAMP ON UPDATE CURRENT_TIMESTAMP,
`type` varchar(45) COLLATE utf8_bin DEFAULT 'mc',
`step` varchar(45) COLLATE utf8_bin DEFAULT NULL,
`host` varchar(45) COLLATE utf8_bin DEFAULT NULL,
`server` mediumblob,
`tempcreds` mediumblob,
PRIMARY KEY (`id`),
KEY `activeIndex` (`active`),
KEY `typeIndex` (`type`)
) ENGINE=InnoDB AUTO_INCREMENT=33888 DEFAULT CHARSET=utf8 COLLATE=utf8_bin;
I am running a query which is mostly based on the table tvnotif which has at least 2 million rows of data in it, all the other tables are having less amount only. I added the index for the tables before that the query took 20 minutes to run and now its taking 160 secs.
EXPLAIN SELECT tvu.secret,COALESCE(php_timezone,"America/Los_Angeles") AS userTz,tn.*,tvu.athleteid,tvu.type FROM tvnotif AS tn
LEFT JOIN vv_us AS tvu ON tvu.token = tn.token
LEFT JOIN tbl_ath_pro AS tap ON tap.athleteid = tvu.athleteid
LEFT JOIN timezones AS tz ON tz.tz_id = tap.tz_id
WHERE tvu.active = 1 AND tn.summary = 'dailies' AND tn.processed = 0
LIMIT 300
Upvotes: 0
Views: 66
Reputation: 48179
The problem is probably your indexes... You have indexes on each field individually. What you need is a composite index on ALL 3 parts as a single index. Without, it can't pick the best one as you have 3 parts of the where clause.
Build a SINGLE index on ( processed, summary, token )
This way the query can jump directly to the processed records, directly to the summary value and then get those records and be done.
Additionally, your VV_US table should have an index on ( token, active ) so the join will be optimized on BOTH parts.
Upvotes: 2