CakePHP: Why is my cached file causing huge spikes when it expires? - cakephp

Question

I'm using cake 2.1.3 and currently have a page that is getting hundreds of views per second and so I have utelized caching in order to handle the load better. The problem is, that once the cache expires, I get a spike in my server resources as well as hundreds of mysql connections.
I'm wondering if I'm going about this the wrong way and if I should be running a cron to cache the page instead of how I'm currently doing it or if there's another technique I'm not thinking of.
here's what my function looks like in my controller:
public function index() {
$this->layout = 'ajax';
if (isset($this->params['url']['callback'])) {
$callback = $this->params['url']['callback'];
}else{
$callback = 'callback';
}
$this->set('callback',$callback);
$today = date("Y-m-d");
$end_date = strtotime ('+1 day' , strtotime($today)) ;
$end_date = date ( 'Y-m-d' , $end_date);
$start_date = strtotime ('-1 day' , strtotime($today)) ;
$start_date = date ( 'Y-m-d' , $start_date);
$total = Cache::read('popular_stories', 'short');
if (!$total) {
$total = $this->TrackStoryView->find('all', array(
'fields' => array('COUNT(story_id) AS theCount', 'headline', 'url'),
'conditions' => array('date BETWEEN ? AND ?' => array($start_date,$end_date)),
'group' => 'story_id',
'order' => array('theCount DESC'),
'limit' => 20,
));
Cache::write('popular_stories', $total, 'short');
}
$this->set('story', $total);
}
Here's what my Cache config looks like in my bootstrap.php file:
Cache::config('short', array(
'engine' => 'File',
'duration' => '+60 minutes',
'path' => CACHE,
'prefix' => 'cake_short_'
));
This is what's in my view file:
<?php
echo $callback . '('.json_encode($story).')';
?>
I was hoping that once the cached file expires, as soon as the first person accessed it, it would craete a new cached file and serve that up for everyone, however because hundreds of people are hitting it per second, it seems like this method isn't working for me and that maybe I should be caching the view view a cron somehow instead or maybe there's a different way to cache that I'm not utelizing.

Answer 1

It sounds like you have the answer more or less figured out (create the cache automatically, not triggered by a user request).
To do this, look into cake's AppShell class, book talks about it here. You can then link this to a cron job. If you create the file thru Cache::write, cake should be aware that it is a new cache file and read it transparently. You might want to leave the "if cache not found" block in there just in case your cronjob fails.
Shells & Tasks in cake are fun and allow you to free your application from using the request/response model exclusively.

Answer 2

TLDR: It's not ideal to force a user to break the cache for you. Use a chron job or a trigger on data change.
Explanation:
"hundreds of views per second" is the problem. When it expires, there are "hundreds of views" during the time it's trying to create the cache file.
The first person hits it, it starts creating the cache, and in the meantime, another hundred+ people hit it, and it looks, and can't yet find a cache file...etc.
If you can manage, try to manually create the cache when an item(s) is updated, or run a chron job that creates a new cache every X minutes as opposed to having it create for a user.
Cake has lots of cool triggers like afterSave() that you can use to trigger this kind of thing. If that doesn't make sense in your case though, a chron job should be fine for you.

Answer 3

I think the answer lies by working out how long this query takes:
$total = $this->TrackStoryView->find('all', array(
'fields' => array('COUNT(story_id) AS theCount', 'headline', 'url'),
'conditions' => array('date BETWEEN ? AND ?' => array($start_date,$end_date)),
'group' => 'story_id',
'order' => array('theCount DESC'),
'limit' => 20,
));
Lets say it takes 500ms.
You are getting 100 hits a second, so when the cache clears the first request makes the find call and then 50 other people also make the find call before the first request completes.
One alternative solution:
Make the cached content never expire. Set up a cron task that overwrites the cache by calling a different action which runs:
Cache::write('popular_stories', $total, 'short');
To overwrite the cached content.
This way, the 100s of users per second will ALWAYS read from cache