Issue with newer version (1.2.16)

[Lalbee99]

I have been using cacti going back to the 0.8.7 days. I have hit bumps alone the road but generally have gotten pointed in the right direction via the forums. The newest version of cacti (1.2.16) has an issue:

CACTI 1.2.16 ISSUE:

1) Less than a day after my Cacti and Spine upgrade (1.2.7 to 1.2.16) I noticed that my snmp based graphs suddenly were not updating. I performed the following troubleshooting steps:
a) I set debugging to Medium - Statistics, Errors, Results. I let a couple polling cycles run and reviewed the cacti.log. There was not a single reference to a snmp based query.
b) I then manually ran spine /usr/local/spine/bin/spine -R -V 3 -S -C /usr/local/spine/etc/spine.conf and it did not perform a single snmp query.
c) I then reviewed the poller cache and there was not a single entry for a snmp query.

Curious note: Any graphs based on the script server worked perfectly.


THOLD 1.4/1.5 ISSUE:

1) I have been using the same snmp based filesystem template with no issues until I recently upgraded the thold plugin from 1.2.1 to 1.4 and also 1.5. Once upgraded to either 1.4 or 1.5 all of a sudden my "Current:" and "High:" values in THOLD which are percentage based (data manipulation) went and became byte based. I tried performing the suggestion cigamit with no effect.

In Example:
BEFORE UPGARDE 35.21 -/80
AFTER UPGRADE 174.32 K 0 / 320.00 K

Curious note: Any template that is script server based honors the Threshold (data manipulation) choice of Percentage


The latest code for both of these has frustrated me. If I am unable to upgrade without loss of functionality I will probably be forced to find an alternative to cacti which I do not want to do.

[bmfmancini]

for the graphing issue

Do you have multiple pollers ?
Try to put a device into debug mode and see what comes up from a poll cycle for that device and work from there

do a select * from poller_output; in MySQL what is showing up in that table ?

[Lalbee99]

I have to no avail. This is clearly a thold issue. Things have worked perfectly until I upgraded to thold 1.5. I have been using Cacti since 0.8.7d and have always managed to work through my issues. On this issue there doesn't appear to be anywhere for me to turn. The developers have told me nope not us. I have thrown up my hands, I have it on my calendar (possibly) 4th qtr to start looking for a replacement.

[macan]

I have to no avail. This is clearly a thold issue. Things have worked perfectly until I upgraded to thold 1.5. I have been using Cacti since 0.8.7d and have always managed to work through my issues. On this issue there doesn't appear to be anywhere for me to turn. The developers have told me nope not us. I have thrown up my hands, I have it on my calendar (possibly) 4th qtr to start looking for a replacement.

any php issue? Which version of php?

[macan]

I had the similar problem. I have error in my script for few years. Tt caused problems in new version of cacti/php. In old version all worked fine even with an error in script . I updated my cacti, add one new device with problematic template and it destroyed maybe 70% of all graphs. I had bits instead of percentage too.

https://github.com/Cacti/cacti/issues/3909

Try to disable all devices except for devices with the one datatemplate, wait few poller cycles. If graphs are ok, enable next devices with another datatemplate and wait few pollec cycles, is it all ok? repeat until you found problematic datatemplate.

[Osiris]

Cacti 1.2.16 has a major performance bug. You should either run the 1.2.x branch or cherry pick the read_config_option() function from 1.2.x. Not too sure about thold.

[Lalbee99]

Currently I poll a client via snmp and retrieve the used and free values for a particular filesystem. The returned values are in 4096 blocks. So naturally within the graph template you would use a CDEF (multiply by 4096) to get the expanded value. Using the CDEF within the graph template causes the value returned in the THOLD tab to change to a metric value (K,M,T) and is definitley not a percentage value. The graphs themselves reflect the proper value for used and free. If I stop using the CDEF in the graph template the THOLD tab shows the current usage percentage however, the graphs show an incorrect value for used and free. Why is it so hard to get someone to look into this. I don't like back rev'ing plugins but clearly there is an issue with THOLD 1.5.

[BAGft]

NB : for new cacti here is the link for weathermap.
sudo wget https://github.com/thurban/plugin_weath ... evelop.zip