[hobbit] Representing disk data for graphing

Francesco Duranti fduranti at q8.it
Thu Oct 12 09:25:44 CEST 2006


> With that I do not understand why I'm getting a graph with 
> the datapoint
>   with one of the items on my status message:
> 
> status vm1.disk green Wed Oct 11 22:12:19 CDT 2006 [vm1] 
> Disk/SFS Utilization OK
> 
> (this is all on one line)
> 
> At this point '11' is a datapoint on the graphs and has a 
> value of zero/

Because the first line have a "/" (Disk/SFS), don't start with "&" and
don't contain "red" or "yellow" (the color of the page is green) so that
line is a "dataline" for do_disk.c.

"11" (from Oct 11) is the 6th field if you count them all from the start
of the line so it's considered the name of the datapoint (corresponding
to the 6th column that is normally the "Mounted on" column.
And 0 is because Oct is not a Number so the conversion function probably
return 0.

Francesco



> 
> Francesco Duranti wrote:
> > No, the problem is not the root filesystem, the problem is that 
> > do_disk.c recognize a disk line if that line contains a "/" 
> inside the 
> > line and will skip that line if it start with "&" and it 
> will skip the 
> > line if it contains the string "red" or "yellow".
> > This is mostly done for Unix filesystem report that looks something
> > like:
> > 
> > yellow  Wed Oct 11 21:55:16 CEST 2006 - Filesystems NOT ok
> > 
> > &yellow /sapdata/QQ8 (94% used) has reached the WARNING level (90%)
> > 
> > Filesystem            kbytes    used   avail capacity  Mounted on
> > swap                 15566048     160 15565888     1%    /var/run
> > swap                 15807304  241416 15565888     2%    /tmp
> > /dev/md/dsk/d0       20645791 6351115 14088219    32%    /
> > /dev/md/dsk/d2       35009161 2241510 32417560     7%    /sapexe
> > 
> > 
> > In this case it need to skip line starting with "&red" or "&yellow"
> > because they are only the alert at the start of the message so that 
> > users know with a direct look what's wrong. Those line have to be 
> > skipped.
> > So a line containing a "/" and not containing a "&" at the start or 
> > the word yellow and red will not be skipped.
> > 
> > I think that you've to change the way you send data because 
> probably 
> > the line with data you send will be skipped if they're reported to 
> > hobbit with "&red" or "&yellow" too.
> > 
> > You can use something like the test I pasted above to send 
> data that 
> > should be recognized as unix filesystem.
> > 
> > 
> >> -----Original Message-----
> >> From: Rich Smrcina [mailto:rsmrcina at wi.rr.com]
> >> Sent: Wednesday, October 11, 2006 9:11 PM
> >> To: hobbit at hswn.dk
> >> Subject: Re: [hobbit] Representing disk data for graphing
> >>
> >> Oh, I see, it's looking for a root filesystem line?  This 
> >> architecture doesn't implement a root filesystem.  In any case I 
> >> would want to just pass over it, reporting and graphing 
> the root is 
> >> as important as any of the other (IMHO).
> >>
> >> Francesco Duranti wrote:
> >>> Not really a problem I think but some misinterpretation by the 
> >>> do_disk.c that is happening if the first line have a "/"
> >>> your first line is:
> >>> "color" 11 Oct 2006 07:45:23 Disk/SFS Utilization OK
> >>>
> >>> I added "color" because it's passed also if you don't see 
> it on the 
> >>> first line because it's the page color.
> >>> The line checking the various line of disk report to get
> >> data line are:
> >>>
> >>>                 /* AS/400 reports must contain the word DASD */
> >>>                 if ((dsystype == DT_AS400) &&
> >> (strstr(curline, "DASD")
> >>> == NULL)) goto nextline;
> >>>
> >>>                 /* All clients except AS/400 report the 
> mount-point 
> >>> with slashes - ALSO Win32 clients. */
> >>>                 if ((dsystype != DT_AS400) &&
> >> (strchr(curline, '/') ==
> >>> NULL)) goto nextline;
> >>>
> >>>                 /* red/yellow filesystems show up twice */
> >>> -->                if ((dsystype != DT_NETAPP) && (dsystype !=
> >>> DT_NETWARE) && (dsystype != DT_AS400)) {
> >>> -->                        if (*curline == '&') goto nextline;
> >>> -->                        if ((strstr(curline, " red ") ||
> >>> strstr(curline, " yellow "))) goto nextline;
> >>>                 }
> >>>
> >>>                 for (columncount=0; (columncount<20);
> >> columncount++)
> >>> columns[columncount] = "";
> >>>                 fsline = xstrdup(curline); columncount = 0; p = 
> >>> strtok(fsline, " ");
> >>>                 while (p && (columncount < 20)) { 
> >>> columns[columncount++] = p; p = strtok(NULL, " "); }
> >>>
> >>> So your dstype is not DT_NETAPP or DT_NETWARE or DT_AS400,
> >> the first
> >>> line have a "/", if the color is not "red" or "yellow" that
> >> line will
> >>> be treated as a data line and it will get some data from it and 
> >>> construct a rrd file.
> >>>
> >>> I had the same problem with netapp filers disk reporting of
> >> my script
> >>> when sending data with the /group option because the 
> /group will be 
> >>> sent to the rrd disk module (you'll not see it in the 
> page but it's 
> >>> there like the color) and in that case I started to have rrd file 
> >>> named with HH:MM:SS ... I got the problem when i started to
> >> get 50-60
> >>> of those files in each filers rrd file.
> >>>
> >>> So it should be safe to skip the first line of the data 
> reported by 
> >>> the do_disk because it will not contain filesystems data
> >> (don't know
> >>> if it can create problem with netware or as400 but it works with 
> >>> netapp/unix/windows clients at least).
> >>>
> >>> Francesco
> >>>
> >>>
> >>>
> >>>> -----Original Message-----
> >>>> From: Rich Smrcina [mailto:rsmrcina at wi.rr.com]
> >>>> Sent: Wednesday, October 11, 2006 8:40 PM
> >>>> To: hobbit at hswn.dk
> >>>> Subject: Re: [hobbit] Representing disk data for graphing
> >>>>
> >>>> Thanks for the patch.  Is there something in the first 
> line itself 
> >>>> that might be causing the problem?
> >>>>
> >>>> Francesco Duranti wrote:
> >>>>> Hi,
> >>>>> One of the problem  is that the first line is got from the
> >>>> do_disk.c
> >>>>> and it have a "/" so it's checked as a data source.
> >>>>> I posted a patch some days ago to solve this problem that
> >>>> also happen
> >>>>> if you use the /group with the bb command. I don't know
> >> if this alo
> >>>>> This will skip the first line of the do_disk message and
> >> it should
> >>>>> then work as expected (at least it work with /group)
> >>>>>
> >>>>> diff -N -r -U5 hobbit-4.2.0/hobbitd/rrd/do_disk.c
> >>>>> hobbit-4.2.0new/hobbitd/rrd/do_disk.c
> >>>>> --- hobbit-4.2.0/hobbitd/rrd/do_disk.c	2006-08-09 
> >>>> 22:10:06.000000000
> >>>>> +0200
> >>>>> +++ hobbit-4.2.0new/hobbitd/rrd/do_disk.c	2006-10-03
> >>>>> 19:58:25.000000000 +0200
> >>>>> @@ -50,10 +50,15 @@
> >>>>>  	else if (strstr(msg, "NetAPP")) dsystype = DT_NETAPP;
> >>>>>  	else if (strstr(msg, "Filesystem")) dsystype = DT_NT;
> >>>>>  	else dsystype = DT_UNIX;
> >>>>>  
> >>>>> - 	curline = msg;
> >>>>> +	/* FD: Skip the first line of message because 
> sometimes it can
> >>>>> create problems with
> >>>>> +		bb /group flag */
> >>>>> +	curline = strchr(msg,'\n');
> >>>>> +	if (curline) curline++;
> >>>>> +	/* FD: End of Skip the first line */
> >>>>>  	while (curline)  {
> >>>>>  		char *fsline, *p;
> >>>>>  		char *columns[20];
> >>>>>  		int columncount;
> >>>>>  		char *diskname = NULL;
> >>>>>
> >>>>>
> >>>>>
> >>>>>
> >>>>>
> >>>>>> -----Original Message-----
> >>>>>> From: Rich Smrcina [mailto:rsmrcina at wi.rr.com]
> >>>>>> Sent: Wednesday, October 11, 2006 8:13 PM
> >>>>>> To: hobbit at hswn.dk
> >>>>>> Subject: Re: [hobbit] Representing disk data for graphing
> >>>>>>
> >>>>>> Oops, I spoke too soon.  It did not work with the icons as
> >>>> I thought. 
> >>>>>> I'm back to no icons and four data points per graph 
> (which looks
> >>>>>> good) and the mysterious third token on the first line
> >> data point.
> >>>>>> Rich Smrcina wrote:
> >>>>>>> Indeed.  I got a suggestion to add the &clear icon to the
> >>>>>> heading and
> >>>>>>> now it works.  Unfortunately I now get one very 
> crowded graph, 
> >>>>>>> otherwise good progress.
> >>>>>>>
> >>>>>>> Thanks for your help!
> >>>>>>>
> >>>>>>> Hubbard, Greg L wrote:
> >>>>>>>> Yes, the "&color" now counts as a token, so any
> >> grep/awk/sed (or
> >>>>>>>> equivalent) is going to be off by one.  Is this
> >>>>>> information included
> >>>>>>>> in the overall data blast from the Hobbit client?  If so,
> >>>>>> you might
> >>>>>>>> consider tacking the colors at the END of each line and
> >>>> see if the
> >>>>>>>> Hobbit parser starts working again.
> >>>>>>>>
> >>>>>>>> For reference, here is something from a Solaris system:
> >>>>>>>>
> >>>>>>>> Wed Oct 11 12:19:50 CDT 2006 - Filesystems OK
> >>>>>>>> Filesystem            kbytes    used   avail capacity  
> >> Mounted on
> >>>>>>>> swap                 18665136  107640 18557496     1%    /tmp
> >>>>>>>> /dev/vx/dsk/rootvol  9210413 1561203 7557106    18%    /
> >>>>>>>> /dev/vx/dsk/var      8263373 1775692 6405048    22%    /var
> >>>>>>>> /dev/vx/dsk/opt2     34291077 2384649 31563518     
> 8%    /opt2
> >>>>>>>> /dev/vx/dsk/home     1021735  614185  346246    64%    
> >>>> /export/home
> >>>>>>>> Notice that the filesystem is token 1, the usage is
> >> token 2, and
> >>>>>>>> available space is token 3, etc., etc.
> >>>>>>>>
> >>>>>>>> These are just ideas.  I've written 7 or 8 custom tests
> >>>> of my own
> >>>>>>>> with my own RRD parsing at the end, so I quickly became
> >>>> acquainted
> >>>>>>>> with the need for agreement between the "pitcher" and the
> >>>>>> "catcher".
> >>>>>>>> GLH
> >>>>>> --
> >>>>>> Rich Smrcina
> >>>>>> VM Assist, Inc.
> >>>>>> Phone: 414-491-6001
> >>>>>> Ans Service:  360-715-2467
> >>>>>> rich.smrcina at vmassist.com
> >>>>>>
> >>>>>> Catch the WAVV!  http://www.wavv.org WAVV 2007 - Green 
> Bay, WI - 
> >>>>>> May 18-22, 2007
> >>>>>>
> >>>>>> To unsubscribe from the hobbit list, send an e-mail to 
> >>>>>> hobbit-unsubscribe at hswn.dk
> >>>>>>
> >>>>>>
> >>>>>>
> >>>>> To unsubscribe from the hobbit list, send an e-mail to 
> >>>>> hobbit-unsubscribe at hswn.dk
> >>>>>
> >>>>>
> >>>> --
> >>>> Rich Smrcina
> >>>> VM Assist, Inc.
> >>>> Phone: 414-491-6001
> >>>> Ans Service:  360-715-2467
> >>>> rich.smrcina at vmassist.com
> >>>>
> >>>> Catch the WAVV!  http://www.wavv.org WAVV 2007 - Green Bay, WI - 
> >>>> May 18-22, 2007
> >>>>
> >>>> To unsubscribe from the hobbit list, send an e-mail to 
> >>>> hobbit-unsubscribe at hswn.dk
> >>>>
> >>>>
> >>>>
> >>> To unsubscribe from the hobbit list, send an e-mail to 
> >>> hobbit-unsubscribe at hswn.dk
> >>>
> >>>
> >> --
> >> Rich Smrcina
> >> VM Assist, Inc.
> >> Phone: 414-491-6001
> >> Ans Service:  360-715-2467
> >> rich.smrcina at vmassist.com
> >>
> >> Catch the WAVV!  http://www.wavv.org
> >> WAVV 2007 - Green Bay, WI - May 18-22, 2007
> >>
> >> To unsubscribe from the hobbit list, send an e-mail to 
> >> hobbit-unsubscribe at hswn.dk
> >>
> >>
> >>
> > 
> > To unsubscribe from the hobbit list, send an e-mail to 
> > hobbit-unsubscribe at hswn.dk
> > 
> > 
> 
> --
> Rich Smrcina
> VM Assist, Inc.
> Phone: 414-491-6001
> Ans Service:  360-715-2467
> rich.smrcina at vmassist.com
> 
> Catch the WAVV!  http://www.wavv.org
> WAVV 2007 - Green Bay, WI - May 18-22, 2007
> 
> To unsubscribe from the hobbit list, send an e-mail to 
> hobbit-unsubscribe at hswn.dk
> 
> 
> 



More information about the Xymon mailing list