Matt: >i remember that i fixed this bug sometime back but it looks like it has >been unfixed. my fix was to simply take the abs() difference in the >timestamps. its simple and results in occasionally getting TN values >that are 1 (or sometimes 2) when they should be 0. ... >gmetad is not happy at all dealing with negative TN values. if you look >in process_xml.c you see tn is defined as a uint32_t.
So Matt is right, gmond should never report negative TN values. In addition to disrupting the logic of gmetad, they do not make any sense. It sounds like the fix is simple. Matt, how is G3? Perhaps we should work on another 2.5.x release with this fix and the xdr_hyper additional type. -Federico
