Every time source one has a value, source two is nan. Every time source two has a value, source one is nan.
nan + value = nan. Therefore you have to set the nan values to 0. CDEF:out0_bits=out0,UN,0,out0,IF,8,* CDEF:out1_bits=out1,UN,0,out1,IF,8,* CDEF:out_bits=out0,out1,+,8,* To see real nan's you could extend this to have it say: if both are nan, the total is nan, else it is the value of the one that is not nan. For a 95th percentile value you would have to alter the script posted here a while back so that it would add both datasources. 95th percentile calculation may be added to the new rrdtool though, I'm not sure.. Serge. -----Original Message----- From: Todd A. Green [mailto:[EMAIL PROTECTED] Sent: Friday, May 23, 2003 6:48 PM To: rrd-list Subject: [rrd-users] Partial data = no graph I have two rrd files. One which was being updated to a certain point then stopped (by choice). The second starting being updated a short time later (by choice). These represent a connection that was moved from one interface to another. The problem is I can generate graphs on the data for the entire month for data sources in either file. I can also generate graphs with different defs for each datasource in the two files into one graph. However which I attempt to generate a combined datasource (CDEF) with the two over the period which the first file wasn't updated I get no graph. I'm pretty sure at this point it's because the first file has not been updated past a certain date. e.g. Both files contain 1 month of 5 minute samples. File 1 updated until March 19th 23:55. File 2 starting updating on March 20th 00:35. Here's a snatch of the RRD scripting DEF:out1=/usr/local/bubba/data/mn/routers/cisco/c7206-1/interfaces/serial2_0/200303.rrd:out:LAST \ DEF:out0=/usr/local/bubba/data/mn/routers/cisco/c7206-1/interfaces/serial1_0/200303.rrd:out:LAST \ CDEF:out1_bits=out1,8,* \ CDEF:out0_bits=out0,8,* \ CDEF:out_bits=out0,out1,+,8,* \ out0_bits & out1_bits would generate a graph, but out_bits would not. Any suggestions. Only thing I can think of is to go back and update the first file to the end of the month with NaN or 0. BTW in looking through the archives I noticed mentions of 95th percentile calcs. That's one of the things my package was designed for. -- 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
