Simon Josefsson wrote: > > I want to monitor things over a long time and I'm worried about > running past the end time of a RRD. > > Should I simply make the RRD end in 50 years? Is there a more elegant > solution?
There's only one solution and that is to *start* it just before the first statistics that you're going to use. Any less and you will lose data, any more is just pure wasting resources. > Currently I'm extracting all data from a SQL database and run rrd > create+update's+graph each time I want a graph, specifying new > start/end times. This is not very elegant but if you need to alter your data after you used them for graphing it's probably the only solution. If you don't need to do so, just update the RRA with the new entries from the SQL database. The rest of this mail isn't valid in that case. You can find out what your earliest data entry is (using SQL) and what the current datetime is. The amount of time involved has to fit in one graph so the next thing to find out is what the amount of time per pixel will be. Set this as the step time. Each CDP in the RRA should then be build from one PDP, run update as you do now and graph the result. This way you should AFAIK get the best performance with the least disk space used. The reason for all this is that the data needs to be in a certain resolution when graphing. It doesn't help to have a better resolution in the RRA as this gets consolidated anyway, it is therefore better to do this at the update stage. cheers Alex -- 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
