I'm playing around with adding DS's to existing RRD files. I was hoping to be able to tell my data collection scripts that if they see new fields, that the .rrd file doesn't have, automaticly take a moment to convert the .rrd file and be done with it.
However, at best, rrdtool dump foo.rrd | rrdtool restore - foo2.rrd takes 15 seconds, consuming 2 CPU's to a maximum (30 CPU seconds). This is without tweaking the XML in the pipe. This .rrd file has 10 DS's, and lots of RRA's (9 RRA's, for a mix of AVERAGE, MIN, MAX, covering various amounts of time and resolution). Now, I know these files aren't small :-) - 3.5 meg per .rrd file. But exporting/important this data via XML is .. "lovely". :-) Has anyone made a way to add_ds using the data structures for rrd files directly? Granted, this is touching the internals of RRD files, and tobias will laugh at me if he changes the data format :-), but I really can't afford to spend 30 cpu seconds (plus filter time) per .rrd file. I have roughly 15,000 .rrd files of similiar size. I intend to take a stab at things, but if someone's already done it.. -- Jason Fesler <[EMAIL PROTECTED]> Good, Fast, Cheap - Pick any two. - RFC 1925 -- Unsubscribe mailto:[EMAIL PROTECTED] Help mailto:[EMAIL PROTECTED] Archive http://www.ee.ethz.ch/~slist/rrd-users WebAdmin http://www.ee.ethz.ch/~slist/lsg2.cgi
