Redis availability check - php

Question

Before using redis I want to check it availability, if redis is not available, I'll use mysql, if redis is available I'll use it. How can I do it, if I use predis client?
My first method was:
/**
* #return bool
*/
public function check(){
if(!#fsockopen ( $server['host'], $server['port'], $errno, $errstr, 3 )){
Debug::instance()->log( 'Redis connect error host: ' . $server['host'] . ' port: ' . $server['port'] );
return false;
}
return true;
}
But it was very bad idea, because I occupy free sockets. Now I try to find a better method.

Answer 1

This should do the trick ;)
// Redis configuration
$vm = array(
'host' => '127.0.0.1',
'port' => 6379,
'timeout' => 0.8 // (expressed in seconds) used to connect to a Redis server after which an exception is thrown.
);
$redis = new Predis\Client($vm);
try {
$redis->ping();
} catch (Exception $e) {
// LOG that redis is down : $e->getMessage();
}
if(isset($e)) {
//use MySQL
} else {
/* Use Redis */
}

Answer 2

Use the predis client and the PING command.
return true if PONG is received as response.
return false on CommunicationException.

Answer 3

Redis can handle far more requests than MySQL, which is why it is so often used as a cache for it. On laptop level hardware I have seen it handling over 1M requests per second - single instance no special server side tuning.
First try to pull the data from redis. If the connection fails, go to MySQL. if you get a null back, go to MySQL and add the data to Redis (optionally with a timeout depending in your criteria) and return the data to the client.
Please do not do a ping before every command. If you have a valid connection, try to get the data and handle getting back nothing. Getting null back means there was no data to pull. Performing the ping before every command is wasteful. You spend a round trip for every call. Often the command to get the data is just or nearly just as fast as the ping, especially considering the cost of testing for the pong. When considering that you will then also be testing to see if you got data returned, the ping before every command model is not wise.
Open your connection in your long term process outside the command, or open the connection, query, and then close your socket each time. I prefer the former, but realize that isn't always possible depending on the framework you are using.
And as someone who has infrastructure with many hundreds of MySQL servers with thousands of DBs, no you can not always assume MySQL is there, just as you shouldn't assume Redis will always be there. Crashes happen, networking happens, servers are tripped up or mistakenly bounced.
But you can follow the flow of "is the TCP connection alive" and "then give me data" followed by validating you got the data. As long as you account for failed connections or timed out requests, and handle them, you are fine.
As far as how to actually code using predis, I recommend starting with the docs such as are found at the Predis github page