external-nagios module should have separate lock file

[Paul Collins]

It's not uncommon to deploy rsync with "max connections" in /etc/rsyncd.conf (nrpe's willingness to clobber this file not withstanding).

"max connections" works by using record locking on the file specified by "lock file".

Both of these are module settings, not global settings, and so when they are set in a global context they set defaults for every module.

By default every module with a connection limit has the same value for "lock file" (the default is /var/run/rsyncd.lock) so they share the available connection slots.

This means that when a limit is set in the standard naive manner, external-nagios is subject to this limit, which causes problems refreshing Nagios from busy Juju units in cases where the script deletes the existing files and fetches them all. (This is done to handle checks that hae been deleted from the Juju unit.)

nrpe should explicitly set "lock file" to something other then the default (I suggest /var/run/rsyncd.external-nagios.lock) and probably set "max connections" as well so that e.g. rate-limiting on swift-storage nodes via "max connections" can't cause problems.