[rrd-users] Fwd: rrdtool resets (sorry)

[Patrick Ammann]

Hi all,

i have some problems in understanding the way to get the "spikes/wraps" out 
of my graphs.
i just read a lot in the mailing list and i finally came to the following 
result.

The setup of my rrd database files looks actually like the following

filename = "test_router.Serial0.rrd"
rrd_version = "0001"
step = 300
last_update = 980849175
ds[in].type = "COUNTER"
ds[in].minimal_heartbeat = 740
ds[in].min = NaN
ds[in].max = NaN
ds[in].last_ds = "2467800446"
ds[in].value = 3.2024009901e+04
ds[in].unknown_sec = 0
ds[out].type = "COUNTER"
ds[out].minimal_heartbeat = 740
ds[out].min = NaN
ds[out].max = NaN
ds[out].last_ds = "1609629600"
ds[out].value = 5.3415594059e+04
ds[out].unknown_sec = 0

1. I could set the "ds[in].min = 0" (as Tobi said). This should cause that 
a negative difference would reslut in a "nan" or "UNKNOWN" value. Did i 
understand this correctly ?

As you wrote, this wouldn't prevent all kind of warps and spikes in the 
graphs.

2. If i change the ds[in].type from "COUNTER" to "Derive" then i should get 
rid off all problems (as discussed in the mail list), right ?

The point that i can't understand is the change of the "ds[in].last_ds". 
Wouldn't this tell to the rrd that the last entry is the number you set to 
that value ? Could you explain to me how this should work and if i realy 
have to set "ds[in].last_ds" to a specific value to get all the warp and 
counter restet problems out of my stats ?








>Original-Recipient: rfc822;Patrick.Ammann at nextra.ch
>Delivered-To: slist at ee.ethz.ch
>Date: Fri, 26 Jan 2001 21:54:03 -0500 (EST)
>From: Christian Pearce <pearcec at PacketPushers.com>
>To: rrd-users at list.ee.ethz.ch
>Subject: [rrd-users] rrdtool resets (sorry)
>X-archive-position: 1364
>Sender: rrd-users-bounce at list.ee.ethz.ch
>X-original-sender: pearcec at PacketPushers.com
>
>
>
>I been doing a lot of reading through the rrd-users archives about counter
>resets that cause rrdtool to artifically wrap.  I don't like rehashing the
>old, but it seems that this topic is an will always cause confusing as
>long as we have counters and rrdtool. :-)
>
>1. Tobi mentions setting the MIN for the DS to 0.  It is my understanding
>when the delta is calculated after a COUNTER was reset it is negative.
>Thereby renedering that interval "nan" or *UNKNOWN*.
>
>2. Check the data on the way in a set it to unknown.
>
>I have tried number one and it didn't work for me.
>
>[bb at yoink larrd]$ rrdtool info
>/usr/local/bbvar/rrd/pearcec.fast.net.apache.rrd
>filename = "/usr/local/bbvar/rrd/pearcec.fast.net.apache.rrd"
>rrd_version = "0001"
>step = 300
>last_update = 980563121
>ds[TA].type = "COUNTER"
>ds[TA].minimal_heartbeat = 600
>ds[TA].min = 0.0000000000e+00
>ds[TA].max = NaN
>ds[TA].last_ds = "998"
>
>I am not going to try and figure out one, but it seems like a good
>solution.  Although from my readings I have been led to believe the MIN
>set to zero isn't a total fix.
>
>So I am just going to use :
>
>ds[TA].last_ds = "998"
>
>as for the last value to compare to my current value.  Is this sane?
>
>And it at all possiable could someone tell me how I might cure number one?
>
>Christian Pearce
>-PacketPusher
>
>
>--
>Unsubscribe mailto:rrd-users-request at list.ee.ethz.ch?subject=unsubscribe
>Help        mailto:rrd-users-request at list.ee.ethz.ch?subject=help
>Archive     http://www.ee.ethz.ch/~slist/rrd-users
>WebAdmin    http://www.ee.ethz.ch/~slist/lsg2.cgi

-------------------------------------------------------------------
Beginner thinks 1Kb == 1000 bytes. Master knows 1Km == 1024m


--
Unsubscribe mailto:rrd-users-request at list.ee.ethz.ch?subject=unsubscribe
Help        mailto:rrd-users-request at list.ee.ethz.ch?subject=help
Archive     http://www.ee.ethz.ch/~slist/rrd-users
WebAdmin    http://www.ee.ethz.ch/~slist/lsg2.cgi