On Wed, Jan 10, 2007 at 12:13:22PM -0500, PNIXON at ci.somerville.ma.us wrote:

On Mon, Jan 08, 2007 at 01:34:30PM -0500, PNIXON at ci.somerville.ma.us wrote:

...

It's the /dev/sdb1 that's not graphing appropriately.

It is graphing on the occasional /mnt that the system does overnight, but not sdb1.

Is there a "disk,snapshot.rrd" file in the ~hobbit/data/rrd/HOSTNAME/ directory ?

--- Yes, the file is there with a current timestamp.

OK, that means it's "only" an issue with getting the graph displayed. Hobbit does pick up the data for this filesystem.

Does the graph show up on the "trends" column ?

--- Yeah, but it's the seventh disk monitored.

Overnight, several remote disks are cifs mounted to be able to be copied locally.

That's the reason. Filesystems that "come and go" tend to confuse the "disk" graph display. I'll bet that your "disk" graph currently has three graphs showing, but it's just not the right three graphs - there are one or more of these remote filesystems showing up instead.

(What's happening is that when building the disk graph, Hobbit looks at the number of filesystems reported in the disk status message. It sees 3 filesystems, so it builds a graph for three filesystems - but when there are more disk*.rrd files than that, it will pick the first three RRD files - in alphabetical order - which were modified within the past 24 hours. Since "mnt" goes before "snapshot", you get a graph for the /mnt filesystem instead of the one for /snapshot).

The best solution would be to put configuration entries for these come-and-go filesystems into hobbit-clients.cfg with an IGNORE setting. Then you can delete the disk*.rrd files that refer to these cifs filesystems. That removes them from the "disk" status, and so they will no longer confuse the disk graph.

Note that this only works if the server is running the Hobbit client (not the BB client).

Regards, Henrik