[Orca-users] Re: Once a day plotting

[Blair Zajac]

luffle wrote:
> 
> Hello,
> 
> I'm trying to plot some daily NetBackup activity. I want to run orca
> only once at noon with only a single line of input everyday. I ran it
> for the past five days with the following input files:
> 
> [zeus]:/usr/local/orca/nblog/input>ls -l
> total 10
> -rw-r--r--   1 root     other         49 Oct  4 16:45 nb.orca.20021004
> -rw-r--r--   1 root     other         49 Oct  5 12:00 nb.orca.20021005
> -rw-r--r--   1 root     other         49 Oct  6 12:00 nb.orca.20021006
> -rw-r--r--   1 root     other         49 Oct  7 12:00 nb.orca.20021007
> -rw-r--r--   1 root     other         49 Oct  8 12:27 nb.orca.20021008
> [zeus]:/usr/local/orca/nblog/input>cat *
> 1033742716       18       88     421948044       88
> 1033812001       32       72     680048925       87
> 1033898400       39       43     584608708       78
> 1033984801       21       21     312067414       69
> 1034072820       13       23     283324056       87
> [zeus]:/usr/local/orca/nblog/input>
> ====================================================================
> I'm running orca with orca -o orcaconf_tape:
> 
> [zeus]:/usr/local/orca/scripts>cat orcaconf_tape
> state_file /usr/local/orca/nblog/state/nb_state
> rrd_dir /usr/local/orca/nblog/rrd
> html_dir /usr/local/orca/nblog/html/orca-html
> html_page_header <a href="index.html">NB INDEX</a>
> generate_hourly_plot    0
> generate_daily_plot     0
> generate_weekly_plot    1
> generate_monthly_plot   1
> generate_quarterly_plot 0
> generate_yearly_plot    0
> 
> group nbjobs {
> find_files /usr/local/orca/nblog/input/nb.orca.(\d{4}\d{2})\d{2}
> column_description etime nbtape nbscratch sumkb nbjob
> date_source column_name etime
> interval 86400
> }
> 
> plot {
> title TAPE USED AND SCRATCH
> source nbjobs
> data nbtape
> data nbscratch
> line_type LINE2
> line_type STACK
> legend cartouches utilisees pendant la nuit
> legend nombre de cartouches dans le scratch
> }
> 
> plot {
> title Kb SUM
> source nbjobs
> data sumkb
> line_type LINE3
> legend kk
> }
> 
> plot {
> title NOMBRE DE JOBS
> source nbjobs
> data nbjob
> line_type AREA
> y_legend nombre de jobs au cours de la nuit
> }
> [zeus]:/usr/local/orca/scripts>
> =================================================================
> The plot I get seem to be some kind of average as the "TAPE USED AND
> SCRATCH" shows a value of 16,333 instead of 13 for the number of
> tapes used last night. What am I missing ? I really can't run my
> script at midnight as my backup window goes from 7 pm to 11am and I'd
> like to update my rrd and html files the same day the backups were
> done (even though some of them actually started the day before).
> Don't pay attention to the type of lines or legends as I am still
> testing (duh!)

That's a tough one.  RRD tool likes to make measurements on integer
multiples of your interval time, so it'll perform some calculation
to place the measurement time at 0:00 GMT time (not local time) since
you've got interval at 86400 seconds.

Maybe the easy solution for you is to print out the time rounded down
or up 86400 when you print your time.

If your script it perl, you'd use something like this:

    my $print_time = int(time()/86400)*86400;
    if (I want tomorrow, not today) $print_time += 86400;

That way you don't have to deal with this.

Best,
Blair

-- 
Blair Zajac <blair at orcaware.com>
Web and OS performance plots - http://www.orcaware.com/orca/