On Fri, Sep 08, 2006 at 08:50:02AM -0500, Scheblein, Adam wrote:
Do you know why they stop graphing??
It's a long story, but what it really boils down to is that there are some silly dependencies between the status column names, the TEST2RRD settings, and the filename of the RRD file being updated. It's a mess, arising from the old BB system which didn't have RRD graphs well integrated, and there is no reason to keep it anymore except that I need to implement something else :-)
The real solution to this is to first split the RRD updates away from the status-column updates; right now, most RRD file updates happen as a result of a status column update (and the data for the RRD file are taken from the status column message text), instead of being triggered by some data that has been collected (e.g. a host load average, the response time of a network service) which is fed directly into the RRD update mechanism. (This also means that Hobbit first spends time collecting all of the data, then generates a text message for the status column, and *then* it spends a lot of effort getting the data *out* of the status message again to put them into the RRD file. Not very clever).
Second, you should be able to "compose" a status message from various bits and pieces of data that Hobbit has. There is no law saying that a "cpu" status column should include the load average data graph. Perhaps it should include some vmstat data graphs instead, or a user-count graph. Or both.
OK, this is getting way off-topic, but - yes, I know why it stops graphing, and I know what it would take to get it right. So it *will* happen at some point :-)
Regards, Henrik