redis-watcher

Redis-Watcher

Watch given keys and metrics on a redis server and regularly reports the results to a statsd server.

Running

redis-watcher -redis=127.0.0.1:6379 -statsd=127.0.0.1:8125

By default this will do nothing; you need to add some keys/metrics to a...

Config file

Basic command reporting

The original use case of redis-watcher was to regularly report the sizes of queues (I use # for comment in the examples; remove them to make a valid .json config file)

{
    # By default, eval each metric every 3 seconds
    "report-interval": 3000
    "metrics": [
        # A simple metric might just be a name, and a command to run
        {
            "name": "crawler.requests.queue.length",
            "command": "zcard all:requests"
        },
        # However you can specify an interval yourself if you like
        {
            "name": "crawler.crawled.queue.length",
            "command": "llen scraped:urls",
            "report-interval": 5000
        }
    }
}

Dump this in /etc/redis-watcher.json and when redis-watcher is run, the zcard all:requests will be reported every 3 seconds.

By default, when a metric is run, the result (which is expected to be a scalar) will be stored using a gauge in statsd. I would be happy to remove both these limitations; I just haven't found a need to yet.

Internal reporting

You can configure redis-watcher to report the results of the INFO redis command. This looks like

{
    "internal": [
        # The key in this entry refers to the key in the
        # datastructure returned by INFO
        {
            "name": "redis.connected_clients",
            "key": "connected_clients"
        }
        
    ]
}