Strange graphs for memory buffer and cached memory

Viruk · Post by **Viruk** » Thu Aug 28, 2008 12:06 pm

Hi,

I have been using cacti for just over a year and it has been working well for both Windows and Linux servers.

I have just set up a new server running Ubuntu 8.04.1 and I have got some strange results.

I have attached screenshots of the graphs produced and a comparison from another server running almost the same setup.

The problem server is running Ubuntu 8.04.1 server and some ticketing software. The server for comparison is running Ubuntu 6.06.1 server and the same ticketing software, although it has a slightly older version installed.

The graphs are from the section "Graph Template: Host MIB - Available Disk Space"

Does anyone have any ideas what might be going on?

I am not sure what other information or logs may be relevant, please let me know what to post if you want to know anything else.

Viruk · Post by **Viruk** » Fri Sep 05, 2008 3:46 am

Sorry just to bump this - but am I missing something obvious?

Viruk · Post by **Viruk** » Mon Sep 29, 2008 4:29 am

This is also happening on another Ubuntu 8.04 server that I have just set up, this isn't happening on my older boxes mainly running Ubuntu 6.06.

Maybe a bug in a later version of a package?

Post by **gandalf** » Wed Oct 08, 2008 12:27 pm

IMHO, there has been a bug fix lately to this template. Please await 087c, appearing soon
Reinhard

Viruk · Post by **Viruk** » Thu Oct 09, 2008 4:03 am

Thanks for your response - I'll look out for it

Viruk · Post by **Viruk** » Mon Jul 27, 2009 6:50 am

I have set up cacti version 0.8.7e on a fresh install of Ubuntu server (8.04.3) and I am still having the same issues with these graphs.

Does anyone have an idea what may be happening? Was the bug resolved in 0.8.7c?

Could there be any packages that I need to update on the servers being monitored?

Post by **gandalf** » Mon Jul 27, 2009 7:01 am

Please verify from log file, that the OIDs cacti is querying are valid for your OS. See 2nd link of my sig for the procedure to do so
Reinhard

Viruk · Post by **Viruk** » Mon Jul 27, 2009 7:05 am

Thanks for a very quick response. I'm checking that out now - the logfile looks ok to me but I'm double checking the kind of things I need to look out for.

Viruk · Post by **Viruk** » Mon Jul 27, 2009 10:02 am

It would appear that I have some NaNs in the .rrd
The NaNs appear in the following lines after using the command:
rrdtool info <rrd file>

ds[hdd_used].max = NaN
ds[hdd_total].max = NaN
rra[0].cdp_prep[0].value = NaN
rra[0].cdp_prep[1].value = NaN
rra[4].cdp_prep[0].value = NaN
rra[4].cdp_prep[1].value = NaN

Is this indicative of invalid OIDs for the OS ?

Post by **gandalf** » Tue Jul 28, 2009 2:06 pm

This _may_ be an indication. But again, see 2nd link ....
It explains step-by-step how to narrow it down
Reinhard

Viruk · Post by **Viruk** » Thu Jul 30, 2009 8:29 am

I've been through everything I can manage on that link:
The log file appears to be fine, everything in debug mode appears to be ok (but that is to my untrained eye). And there are no errors running at normal logging mode.

The poller is running correctly and crontab shows poller.php

The rrd files are being created ok and graphs are all visible.

All the rrd files are owned by the cactiuser, not by root

As I mentioned before I am seeing NaNs in some of the values when checking rrds with rrdtool info, but I am not sure what these values should be instead and I'm not confident with the use of rrdtool tune

php.ini has been changed so that the memory setting is 256MB instead of the default 8MB limit (I used to use 128MB on my old cacti installation but increased due to better hardware running the new install)

I'm struggling to identify which part of the debugging section is applicable to verifying that the OIDs are valid for the OS - is there a specific section I should be paying attention to there?

I am going to have a look around for an updated MIB as something has changed between Ubuntu 6.06 and 8.04, I just can't tell what it is.

Post by **gandalf** » Thu Jul 30, 2009 1:19 pm

I would need the output of the SNMPwalks against the memory OIDs as a start to judge what data comes back from the target.
Reinhard

Viruk · Post by **Viruk** » Fri Jul 31, 2009 9:42 am

I ran:
snmpwalk -Os -c [community] -v 1 [target IP] memory

I hope that's what you're looking for!

The output is as follows:
memIndex.0 = INTEGER: 0
memErrorName.0 = STRING: swap
memTotalSwap.0 = INTEGER: 851960 kB
memAvailSwap.0 = INTEGER: 851880 kB
memTotalReal.0 = INTEGER: 515384 kB
memAvailReal.0 = INTEGER: 52932 kB
memTotalFree.0 = INTEGER: 904812 kB
memMinimumSwap.0 = INTEGER: 16000 kB
memShared.0 = INTEGER: 0 kB
memBuffer.0 = INTEGER: 64764 kB
memCached.0 = INTEGER: 338880 kB
memSwapError.0 = INTEGER: noError(0)
memSwapErrorMsg.0 = STRING:

Post by **gandalf** » Fri Jul 31, 2009 2:59 pm

And which of those OIDs is used by the faulty graph?
Reinhard

Viruk · Post by **Viruk** » Mon Aug 03, 2009 4:32 am

I think that it is:
memBuffer.0 = INTEGER: 64764 kB
memCached.0 = INTEGER: 338880 kB

Cacti

Strange graphs for memory buffer and cached memory

Strange graphs for memory buffer and cached memory

Who is online