Some Strange things happen after I change the rra setting

[seraphita]

I install a new cacti server and changed the rra setting at the following:

1. I changed the daily rra's rows from the default 600 to 18000
2. I changed the weekly rra's rows from the default 700 to 18000
3. I changed the monthly rra's rows from the default 775 to 9000 and the steps from the default 24 to 12

then something strange happened,when i see the same interface's traffic on this cacti server and my old cacti server, i found that the traffic on the new server is more higher than the old one.
on the old server the traffic is 100Mbit/s but on the new server it is 150Mbits/s!!
Is there anything wrong when i change the rra settings?

[BSOD2600]

No. Old rrd files will not have their row values changed (unless you manually do that). Only new rrd files will use the larger row values.

When you say 100Mbit/s vs 150Mbit/s, you're talking about the actual bandwidth or just the graph axises? Post some pics.

[seraphita]

the real.png is the real traffic of the port
the test.png is the graph created by new cacti server

And on the new server, the graph "test" is created after i change the rra setting. I also check the rra files ,the filesize is up to 4MB. I think the file size should be right.

I try to create the graph manually and get the same graph
[/img]

[seraphita]

the following is the rra setting on my new cacti server
And I only change the rra setting on the new server

[gandalf]

See somem of the first links of my signature for rrdtool related topics such as rra definitions and the like
Reinhard

[seraphita]

Thank you for your answer
I have read your post
" Howto define very BIG rra "without data loss" "
http://forums.cacti.net/viewtopic.php?t=9605

and follow the steps and create a new graph
but i got the same wrong graph

[gandalf]

First, rra settings will not modify any data put into the rrd file. RRA defines the way of consolidation only.
For debugging, please switch settings -> Logging Level to DEBUG and perform some runs on both machines in parallel. You'll find all snmpget commands accompanied with the results in log/cacti.log. Find out the datasources pertaining to your graphs and compare the values from the snmpget.
But pay attention: Traffic OIDs are COUNTER types, so the absolute values are not of interest but the differences between two pollings. If this gives you the same differences, I suppose that this is the real difference between what both interfaces are seeing. Perhaps some tcpdump/ethereal tracing will give insight
Reinhard

[seraphita]

I think the problem I meet maybe because the cacti-0.8.6h on debian has some problem.
When I do the same test with cacti-0.8.6c on debian
or test with cacti-0.8.6g on redhat
they all work fine

[gandalf]

I don't know anything about those distros. Cacti developers feel themselves responsible for the code they distribute on the main cacti site only. Personally, I'd recommend the src.rpm for rpm based distros (as this is all php/java, no compile needed!).
Reinhard