scalefactory / zcollective Goto Github PK

On a machine running zcollective as a cron job, but where no activemq is present, zcollective keeps running and retrying. Ensure timeouts exist properly and/or use a lock file to ensure only one copy runs at a time.

If a machine running zcollective is out of disc space, the lock file can be written but left at zero bytes. Be more robust in zcollective about handling invalid lock file state.

In L373-385 of bin/zcollective we filter the hosts that mcollective has discovered to only include the ones which match the given options[:interface_cidr] if set.

For performance reasons, should we also be filtering the hosts that we get out of Zabbix if this is set and only deal with hosts in there that match the IP range? ie. we would skip L299-314 if it's out of the given range?

A "verbose" output level would be very useful. This could be the same as debug, but omitting the HTTP requests and responses.

It would be useful and possibly more secure if zcollective could be configured to read it's access credentials for zabbix etc. from a configuration file rather than being given them as command line parameters.

Make it easier to test and release this gem by adding Travis configs.

The current hostgroup membership assignment of nodes only occurs during the initial import, and not on subsequent runs; as a consequence, hosts won't be added to new subcollectives that have been added.