trinitum
commented
10 years ago The on_connect_error callback allows you to change host/port in case connection to redis has failed. Default on_connect_error callback confesses. If callback does not die, then reconnection loop continues indefinetely. If you want to have just one connection to any of the three servers then you should use this callback:
use 5.010;
use strict;
use warnings;
use RedisDB;
my @ports = (6371,6372,6373);
my $redis = RedisDB->new(
port => $ports[0],
on_connect_error => sub {
push @ports, shift @ports;
warn "Disconnected from $_[0]->{port}, connecting to $ports[0]";
$_[0]->{port} = $ports[0];
},
);
while (1) {
my $len = $redis->llen("queue_opsstats");
printf "%d\n", $len;
sleep 1;
}In your case you seems want to read from all the servers. Set raise_error to false, and do not define on_connect_error callback. In this case when connection fails redis operation ('llen' in your case) will return RedisDB::Error::DISCONNECTED object, based on that you can skip reply from the server.
use 5.010;
use strict;
use warnings;
use RedisDB;
use Scalar::Util qw(blessed);
my @queues;
foreach my $i ( 6371, 6372, 6373 ) {
my $redis = RedisDB->new(
host => "redis",
port => $i,
raise_error => 0,
);
push @queues, $redis;
}
my $last;
while (1) {
my ( $total, $len ) = 0;
print "len:";
foreach my $redis (@queues) {
my $len;
$len = $redis->llen("queue_opsstats");
if ( blessed $len) {
# error
}
else {
$total += $len;
printf " %d", ( defined($len) ? $len : -1 );
}
}
printf " (%d), queue diff: %d\n", $total, $total - $last;
$last = $total;
sleep(1);
}Note that each time you send new request to redis it will try to reconnect, which blocks for some time, so you may want to stop sending requests to this server for a couple of minutes before giving it another chance, alternatively use some AnyEvent based module, e.g. AnyEvent::Redis.
mburgoon
I'm using Redis as a queue utilizing lpush and [b]lpop to get items in and out of the queue. I'm going to be running several instances of redis to provide high availability, scalability as well as redundancy on multiple machines.
However, with RedisDB, I am unable to get it to break out of the redis call upon server failure. For this module to work for me, I need to be able to detect an error when talking to a redis server, mark that connection as bad so that I will not use it for some time, and then try to connect back to that server later.
Here's the basic script that I'm using to validate behavior. Regardless of what I put in raise_error, reconnect_attempts, etc, the RedisDB module never returns until the redis server that I stopped comes back online. Whenever I stop an instance of redis, I just get a bunch of dots signifying that the on_connect_error callback was being executed over and over again.
Is there something basic that I'm missing or an approach that I haven't seen?