Only half of the data from an interface poll?

[gandalf]

Ok. I suppose I can't help finding the reason for those counters giving such strange throughputs. Please consider asking the vendor
Reinhard

[TheWitness]

Here is the "maximum" 64-bit unsigned integer value:

9223372036854775807

This number is 19 digit's long. Therfore, the number posted before:

See the number 456441259364798423? Isn't it a bit huge? So are all of them! So rrdtool clips them off during update, as they exceed the defined rra's maximum.

.

It is only 18 digit's long, one order of magnitude below overflow. But here is a thought. First of all, if you run the following:

Code: Select all

rrdtool info <my_rrd>

Is your max value set? Is it high enough? You can use the following command to clean it up:

Code: Select all

rrdtool tune [options] <my_rrd>

This could also an issue with your build of rrdtool. Is it a 64-bit build? Does this make a difference to rrdtool? Maybe, after you verify the information from above and "possibly" discover the issue here, you can post to Tobi's forum and then post your response here.

TheWitness

[gandalf]

Here is the "maximum" 64-bit unsigned integer value:

9223372036854775807

This number is 19 digit's long. Therfore, the number posted before:

See the number 456441259364798423? Isn't it a bit huge? So are all of them! So rrdtool clips them off during update, as they exceed the defined rra's maximum.

.

Yep, I know. And indeed, given the "rrdtool create" statements taken from the log, rrdtool clips the values (not because they are huge in absolute figures but because the COUNTER differences between two of them are huge. I only was too lazy to write all of this down)
So I will talk about all steps I've performed
1. Took the rrdtool create statements from the first log
2. Removed the MAXimums in the create statements and created those rrd files (I had to add a --start option to let those data in)
3. Then I sorted all data given in the logs and fed them into this new one
4. I created a standard Traffic Graph and related it with those rrd files
5. I received the graphs shown above. Please find, that Gi1.1 is reported to have more than 1 TB input. This is indeed a bit huge
If there is an error in my approach, please let me know.
Reinhard

[TheWitness]

I am guessing that this is an issue related to 64bit counters and a 32bit RRDtool build. I think the question has to be asked of Tobi and crew.

If RRDtool built 64-bit and if so, has this been confirmed against both versions?

TheWitness