[Nagiosplug-devel] additional function for check_disk [Virus checked]
Andreas Ericsson
ae at op5.se
Tue May 17 10:55:09 CEST 2005
Matthias.Eble at kaufland.de wrote:
> Hi list,
>
> at the moment, check_disk returns a state != STATE_OK if any of the
> monitored filesystems has less than -w or -c freespace.
>
> In our case, this doesn't meet our needs, since our Oracle databases use
> more than just one filesystem to store their data.
> So if there are n data-volumes available, it doesn't matter if one or n-1
> volumes get filled 100% as long as one volume remains beyond
> the warning level.
>
> I've got three approaches for solve this problem.
>
> First:
> Introduce an option like -g (--group) followed by a comma_seperated_list
> of partitions/lvols/mountpoints which are to group.
>
Probably the cleanest solution, or possibly to add check_disk_group as
its own command.
>
> Second:
> Use a own nrpe command for every group of filesystems and introduce a flag
> like -g (in addition to multiple -p options) to tell check_disk to return
> warn/error if all
> filesystems monitored by this command get full
>
> Third:
> Build an ugly shellscript which executes check_disk for every
> datafilesystem and return != 0 if all of these checks return !=0.
>
Shell is good for prototyping, but the functionality would be equivalent
to that of check_cluster (or check_cluster2, I can't remember).
>
> A further parameter could be sth like minimum_filesystems_ok to decide if
> the state is ok if 1,2,3,... filesystems of a group have more than -w/-c
> free space.
>
>
> What do you think about it?
Hack away. A clean patch is almost certain to be included upstream.
> Is the check_disk plugin is the right place to reproduce a functionality
> like this.
>
> Does anybody know applications like Oracle which use more than one
> filesystems for the same purpose?
>
Samba and various ftp servers in different setups. I'm sure there are
more scenarios where this can be interesting.
--
Andreas Ericsson andreas.ericsson at op5.se
OP5 AB www.op5.se
Lead Developer
More information about the Devel
mailing list