Cassandra allow checking for up/down nodes from multiple hosts.

[dmilcevski]

This pull-request allows for checking multiple Cassandra hosts in a cluster in a case where one host is down, the other hosts might provide the information how many nodes are up or down. If you check only one host, and it happen to be that this host is down, you cannot get notifications. The fixes allow for checking for multiple addresses, and if all hosts in the set are down, then a CRITICAL message is returned.

This pull request also requires changes in the Nodetool.pm file. I will create a separate pull request for this. The changes there are to allow skipping connection refused from down hosts, but still able to return OK if the number of up hosts is bigger then the threshold.

[HariSekhon]

Thanks very much for the pull requests.

This can actually be solved in a simpler and more generic way by using either a Load Balancer (HAProxy is free and config is ready-made and available below, which is also a sub-repo to this one and used in many CI tests):

https://github.com/HariSekhon/haproxy-configs/blob/master/cassandra-jmx.cfg

or via a dynamic query to find_active_server.py in a subshell to pass to any Nagios Plugin, so you don't have to rewrite and complicate existing Nagios Plugins for technologies that later added Active/Passive Master HA like Hadoop etc.

https://github.com/HariSekhon/devops-python-tools/blob/master/find_active_server.py

The load balanced method is better in that it reduces the number of queries being sent by tools like Nagios Plugins when one or more of the nodes are offline (it's common on larger clusters to have some nodes offline for maintenance, disk replacements, patches, upgrades etc).