[rrd-users] logging old data.... RRD update frequency....
Daniel A. Murray
murdan at cox.net
Thu Apr 24 12:36:27 MEST 2003
does anyone know how to log "old" data? I keep erroring out that the
"update" time is prior to the time of the last update.
I am gathering execute time statistics. Every 5 minutes I start a process.
This process generally takes less than two minutes to complete. When the
process is complete, a results file is saved that includes the start and
stop time of the process.
The script that starts the process also checkes to see if the results file
exists, and if so, parses it and then passes it to RRD with the start and
elapsed time as parameters.
While this generally works ok, on occasion a wayward process sometimes takes
longer than 5 minutes, and occasionally as long as 48 hours. The problem is
how to accurately go back and log that data which, from an RRD pespective
would have been missed.
Any thoughts? The only way I can think of is a brute force method of
exporting the entire RRD, inserting the data, then re-importing the entire
RRD. There has to be a better way.
----- Original Message -----
From: "Bas Rijniersce" <bas at brijn.nu>
To: <rrd-users at list.ee.ethz.ch>
Sent: Wednesday, April 23, 2003 11:23 AM
Subject: [rrd-users] Wrapping counter
> I use RRDtool to monitor some things on my home server. One of the items
> is the Apache webserver statistics. I used the following RRD
> ... DS:totxs:COUNTER:600:U:U DS:totkb:COUNTER:600:U:U ...
> RRA:AVERAGE:0.5:1:288 RRA:AVERAGE:0.5:6:336 RRA:AVERAGE:0.5:12:744
> RRA:AVERAGE:0.5:288:730 RRA:MAX:0.5:1:288 RRA:MAX:0.5:12:744
> The totxs DS is the counter of number of accesses logged by apache, if
> Apache is restarted, this counter returns to 0. The RRD manual says:
> "The COUNTER data source assumes that the counter never decreases,
> except when a counter overflows. The update function takes the overflow
> into account"
> I guessed that this intelligent wrapping would also be true for wraps
> like I described above. Unfortunately that didn't work for me. Have a
> look at the apache week graph at http://www.brijn.nu/server
> A hundred million hits is not something my server would survive :-).. As
> a result the significant part of the graph is scaled out of sight.
> What is the correct way to make sure this doesn't happen again??
> 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
Unsubscribe mailto:rrd-users-request at list.ee.ethz.ch?subject=unsubscribe
Help mailto:rrd-users-request at list.ee.ethz.ch?subject=help
More information about the rrd-users