I am monitoring a set of load balancers (Alteons) which each have at least 1, but possibly/usually more, virtual addresses configured (for server load balancing groups). At the moment, I use a custom script to poll each device via SNMP and send a single status message per device to Hobbit, based on the session loads on those "virts." I am using NCV for graphing.
However, with this, I am only able to manage one graph, and one data source, per device. I would like to graph all the session values. I would also like to use multiple RRD files. My experience with the single-file, multiple-DS approach is that, if I add another virtual, graphing will break (updates error out, probably since the RRD was not created with this new data source defined).
I am amenable to changing the scripts to use the data channel approach (using the --extra-script method). I am hoping that I could use the hobbitgraph RRDFN/RRDIDX/RRDPARAM keywords to create the graphs. But, the documentation for custom graphing implies that only a single filename for the incoming data can be specified, so I see no way to split the data into multiple files, single DS per RRD.
Short of creating a separate test column for each virtual I am monitoring, is there a better way to collect/plot this information? Thanks in advance for any suggestions. -Alan
-- Alan Sparks, UNIX/Linux Systems Integration and Administration <asparks at doublesparks.net>