cacti graphing wrong data

anervoustwitch · Post by **anervoustwitch** » Thu Aug 28, 2008 1:39 pm

cacti is running on a linux server, graphing (among other devices) two linux machines.

cacti is graphing the data from one of the machines in both graphs and ignoring the data from the other machine.

DNS is correct, i checked the snmp settings and neither is acting as an agent for the other so theres nothing funky there.

anyone have an idea what might be going on?

Post by **gandalf** » Fri Aug 29, 2008 3:29 pm

When editing both devices, what is displayed on the upper left?
Reinhard

jesterx · Post by **jesterx** » Mon Sep 01, 2008 2:58 am

you've probably set the device type to 'local linux machine' in which case it's just graphing stats from the local cacti box for both machines. i think the wording there is a bit misleading, it caught me out as well the first couple of days or so.

bwallach · Post by **bwallach** » Thu Oct 23, 2008 11:52 am

Did you ever solve this? It looks like I am having the same exact issue.

We have 2 linux boxes we wish to monitor. Cacti is installed on one of them. snmpd is running on both. We have set up graphs (using the standard unix/linux templates) to monitor mem/load/users/processes/apache hits.

The localhost graphs show data that appears to be correct. However, the remote host graphs are showing the localhost data?

Note, we did try changing the device Host Template "host type" to "None", "Generic SNMP-Enabled Host", & "ucd/net SNMP host" options (as well as initially "Local Linux Machine". None of these seem to make a difference though.

Thanks in advance for any insights!

Post by **gandalf** » Thu Oct 23, 2008 2:05 pm

If the remote host was added as "local unix host" in the first instance, changing to host template "none" does not make any difference. Please delete the host and re-create using the correct (e.g. ucd/net) host template.
Reinhard

bwallach · Post by **bwallach** » Thu Oct 23, 2008 3:49 pm

Thanks for your reply Reinhard.

We deleted the host and created a fresh one set to "ucd/net". We then added some of the standard ucd/net graphs for disk/cpu/load/mem (for this new host). The graphs do not show any data though?

We added some of the standard graphs for unix load/users/pings/processes (for the remote host again) and these do show data, but once again it is the localhost data that is being rendered in the graphs.

I suspect that we are doing something fundamentally wrong here?

Thanks again

Post by **gandalf** » Fri Oct 24, 2008 6:40 am

What happens, if you snmpwalk against that target host?
Reinhard

bwallach · Post by **bwallach** » Fri Oct 24, 2008 11:04 am

This is the snmp.conf file on the remote host:
----------------------------
rocommunity public ip_address_of_cacti_host
rocommunity public localhost

syslocation PROD1
syscontact my_contact_email

dontPrintUnits true

includeAllDisks

proc httpd
----------------------------

ran the following queries from cacti host machine (per snmp tutorial):
------------------------------------------------
# snmpwalk -v 1 -c public ip_address_of_remote_host .1.3.6.1.2.1.1.1.0

response:
SNMPv2-MIB::sysDescr.0 = STRING: Linux prod1 2.6.17-1.2174_FC5 #1 Tue Aug 8 15:30:55 EDT 2006 i686

# snmpwalk -v 1 -c public ip_address_of_remote_host .1.3.6.1.4.1.2021.100.2.0

response:
UCD-SNMP-MIB::versionTag.0 = STRING: 5.3

# snmpwalk -v 1 -c public ip_address_of_remote_host .1.3.6.1.4.1.2021.9.1.2

response:
nothing

# snmpwalk -v 1 -c public ip_address_of_remote_host .1.3.6.1.2.1.2.2.1.2

response:
IF-MIB::ifDescr.1 = STRING: lo
IF-MIB::ifDescr.2 = STRING: eth0
IF-MIB::ifDescr.3 = STRING: sit0
-------------------------------------------------

Post by **gandalf** » Sat Oct 25, 2008 1:09 pm

bwallach wrote:Thanks for your reply Reinhard.

We deleted the host and created a fresh one set to "ucd/net". We then added some of the standard ucd/net graphs for disk/cpu/load/mem (for this new host). The graphs do not show any data though?

We added some of the standard graphs for unix load/users/pings/processes (for the remote host again) and these do show data, but once again it is the localhost data that is being rendered in the graphs.

I suspect that we are doing something fundamentally wrong here?

Thanks again

You will have to use the ucd/net stuff for the remote host. Using the "other" graphs will always poll localhost only.
So please again adjust to those graphs only.
Then, please tell us, which downed host detection is used for that host and if snmp data is shows on the upper left (please post the data). Then, please tell us the host's status (use the "Device" list).
Next: are you using cmd.php or spine?
Reinhard

bwallach · Post by **bwallach** » Sat Oct 25, 2008 4:58 pm

I got rid of the unix/linux graphs...just using ucd/net graphs (and no data showing)

using poller.php for cron on cacti host machine which looks like is using cmd.php (I did not explicitly install spine)

Status for remote host device is "Up"

SNMP Information from console->device view for remote host device:

System: prod1 2.6.17-1.2174_FC5 #1 Tue Aug 8 15:30:55 EDT 2006 i686
Uptime: 11147207 (1 days, 6 hours, 57 minutes)
Hostname: prod1
Location: PROD1
Contact: username@myemail.com

Post by **gandalf** » Wed Oct 29, 2008 2:06 pm

bwallach wrote:I got rid of the unix/linux graphs...just using ucd/net graphs

fine

using poller.php for cron on cacti host machine which looks like is using cmd.php (I did not explicitly install spine)

Ok

Status for remote host device is "Up"

Ok

SNMP Information from console->device view for remote host device:

System: prod1 2.6.17-1.2174_FC5 #1 Tue Aug 8 15:30:55 EDT 2006 i686
Uptime: 11147207 (1 days, 6 hours, 57 minutes)
Hostname: prod1
Location: PROD1
Contact: username@myemail.com

Ok

Let's only address prod1 for now. Please visit "System Utilities -> View Poller Cache"; filter for prod1 and post the results found (replace snmp community strings before posting the screenshot).
We will next try to use snmpwalk manually to verify snmpd installation on prod1
Reinhard

bwallach · Post by **bwallach** » Thu Oct 30, 2008 3:34 pm

When I filter on prod1, nothing appears on the poller cache screen.

When I remove the filter, I see various entries for the prod2 unix graphs though...like this:

PROD2 - Apache- Hits Script: perl /usr/share/cacti/scripts/webhits.pl
RRD: /usr/share/cacti/rra/prod2_hits_20.rrd
PROD2 - Load Average Script: perl /usr/share/cacti/scripts/loadavg_multi.pl
RRD: /usr/share/cacti/rra/localhost_load_1min_5.rrd

Post by **gandalf** » Fri Oct 31, 2008 4:20 pm

But you already did create new graphs for prod1?
Reinhard

bwallach · Post by **bwallach** » Fri Oct 31, 2008 8:33 pm

Yes, there are 4 prod1 ucd/net graphs & 1 prod1 unix ping latency graph (I left that one around).

See graph management screenshot attached.

Post by **gandalf** » Mon Nov 03, 2008 4:13 pm

Can't make sense of it. Is online access to your system possible? Please pm details, if possible
Reinhard

cacti graphing wrong data

cacti graphing wrong data

Who is online