Solaris 9 problem with snmp graph

[bowbow]

I turned on cacti debug, and I see the snmp query for disk usage works, also the png file has a new time stamp. I am using net-snmp on the remote host. When I run an snmpwalk manually the type returned is INTEGER.

01/07/2006 12:36:11 AM - CMDPHP: Poller[0] Host[29] DS[244] SNMP: v1: HOSTNAME, dsname: dskPercent_var, oid: .1.3.6.1.4.1.2021.9.1.9.2, output: 34
01/07/2006 12:36:11 AM - CMDPHP: Poller[0] Host[29] DS[243] SNMP: v1: HOSTNAME, dsname: dskPercent_export, oid: .1.3.6.1.4.1.2021.9.1.9.3, output: 3
01/07/2006 12:36:11 AM - CMDPHP: Poller[0] Host[29] DS[242] SNMP: v1: HOSTNAME, dsname: dskPercent_root, oid: .1.3.6.1.4.1.2021.9.1.9.1, output: 64

The rdtool runs, the debug log shows it. I even ran the rdtool command manually and incremented the time stamp by 1 second.

01/07/2006 12:36:11 AM - POLLER: Poller[0] CACTI2RRD: /usr/local/bin/rrdtool update /usr/local/apache2/htdocs/cacti/rra/HOSTNAME_diskroot_242.rrd --template dskPercent_root 1136615762:64
01/07/2006 12:36:11 AM - POLLER: Poller[0] CACTI2RRD: /usr/local/bin/rrdtool update /usr/local/apache2/htdocs/cacti/rra/HOSTNAME_diskexport_243.rrd --template dskPercent_export 1136615762:3
01/07/2006 12:36:11 AM - POLLER: Poller[0] CACTI2RRD: /usr/local/bin/rrdtool update /usr/local/apache2/htdocs/cacti/rra/HOSTNAME_diskvar_244.rrd --template dskPercent_var 1136615762:34

To no avail, running the strings command against the rrd file shows no last value added to the rrd file so the graph is not generated as it should be. Even though the snmptrap shown above it returning the values 34, 3, and 64. The Data Source type of Gauge was used in cacti. Again the snmpwalk type is INTEGER.

here is the output of strings:

strings /usr/local/apache2/htdocs/cacti/rra/HOSTNAME_diskroot_242.rrd
0003
dskPercent_root
ABSOLUTE
AVERAGE
AVERAGE
AVERAGE
AVERAGE
LAST
LAST
LAST
LAST
UNKN


Can someone please help with what I am doing wrong?

[gandalf]

Please run the rrdtool fetch command against the rrd to see the last updated values
Reinhard

[bowbow]

I've noticed that it seems to be a problem when the snmpwalk command responds with INTEGER: When the snmpwalk return value includes COUNTER32: or STRING:, the graph is created without a problem. I saw somthing in the rrdtool fetch man pages about printf. Could there be somthing there that I could test?

/usr/local/net-snmp/bin/snmpwalk -Of -v 1 -c COMMUNITY HOSTNAME OID
returns the following:

/var snmp trap
.1.3.6.1.4.1.2021.9.1.9.2
.iso.org.dod.internet.private.enterprises.ucdavis.dskTable.dskEntry.dskPercent.2 = INTEGER: 34

/export snmp trap
.1.3.6.1.4.1.2021.9.1.9.3
.iso.org.dod.internet.private.enterprises.ucdavis.dskTable.dskEntry.dskPercent.3 = INTEGER: 3

rrdtool fetch HOSTNAME_diskexport*rrd AVERAGE -r 1 -s -1h
dskPercent_export

1137341700: 5.0161202186e-02
1137342000: 5.0000000000e-02
1137342300: 4.9833333333e-02
1137342600: 5.0000000000e-02
1137342900: 5.0000000000e-02
1137343200: 5.0166666667e-02
1137343500: 5.0000000000e-02
1137343800: 5.0000000000e-02
1137344100: 5.0000000000e-02
1137344400: 5.0000000000e-02
1137344700: 5.0000000000e-02
1137345000: 5.0000000000e-02
1137345300: NaN

rrdtool fetch HOSTNAME_diskvar_244.rrd AVERAGE -r 1 -s -1h
dskPercent_var

1137341700: 5.6849362477e-01
1137342000: 5.6666666667e-01
1137342300: 5.6477777778e-01
1137342600: 5.6666666667e-01
1137342900: 5.6666666667e-01
1137343200: 5.6855555556e-01
1137343500: 5.6666666667e-01
1137343800: 5.6666666667e-01
1137344100: 5.6666666667e-01
1137344400: 5.6666666667e-01
1137344700: 5.6666666667e-01
1137345000: 5.6666666667e-01
1137345300: NaN

The MAX, MIN, and LAST values are very similar

example:
rrdtool fetch HOSTNAME_diskexport*rrd LAST -r 1 -s -1h

1137342300: 4.9833333333e-02
1137342600: 5.0000000000e-02
1137342900: 5.0000000000e-02
1137343200: 5.0166666667e-02
1137343500: 5.0000000000e-02
1137343800: 5.0000000000e-02
1137344100: 5.0000000000e-02
1137344400: 5.0000000000e-02
1137344700: 5.0000000000e-02
1137345000: 5.0000000000e-02
1137345300: 5.0000000000e-02
1137345600: 5.0000000000e-02
1137345900: NaN

Just some info. My goal is a cacti implementation to monitor 120 Solaris servers, and a linux/windows blade server farm. Figuring out the Solaris monitor is just step one.

Thank you for your help lvm, any more help would be very much appreciated.

[bowbow]

This is what I had to do to obtain the snmp values. I moved the data source type of gauge from the template to a per item value which fixed the problem.

These are the steps I used.

1. Change the Data Template:
Check the box for Data Source Type so its set to "Use Per-Data Source Value.
2. Create a New Data Source:
Use the updated data template: Use the value Gauge for Data Source type which before, was being provided by the template.
3. Change the Graph Management:
Use the new data source in the graph of choice that I just created. Bang! no more NaN. Now I have values instead and the graphs are updating.
4. Delete the old Data Source:
Save on confussion in the future when you don't remeber which one was working

Hope this helps someone else out some day.

Strange, but it worked.