what I meant is that if the true fan readings are usually 4111,4221, and 4309, for example, you'd expect to see a "stairstep" type of plot where only those values are shown and all the lines are vertical or horizontal; if there were interpolation, however, intermediate values would be plotted, which would make it appear that the fan sensor had more resolution than it actually did. Jean Delvare wrote: >>even if he increases the fan divisor so he doesn't get 0 anymore, >>fan readings don't have a lot of resolution, so the values >>might be 4111, 4221, 4309, etc., so I can see where >>interpolation could be misleading. > > > I don't. Who cares about the exact reading (which anyway is already a > truncated value)? If rrd averages 4111 and 4221 to, say, 4139, well, > this is a relevant value (while averaging 4111 and 0 to, say, 300 > wouldn't). > > As a side note, the 0 reading cannot be due to a bad fan divisor. Mario > set the divisors to 2 and 4200 is no way near the limit for an AS99127F > (it is below 3000 if I remember correctly). More likely the fan is > broken > > I also think that it is possible to tell rrd to ignore values outside of > a given range. You could make it so that 0 isn't in the range, so it be > taken into account. However, this makes monitoring the fan rather > useless, of course. In any case I would consider buying a new, working > fan. Trying to work around buggy hardware through software tricks rarely > works (and is never the right thing to do). >
