better ucd/net-snmp memory usage -- update June 21 2009

[sauron]

By all means compile it in. The downsides are minimal, and you get another 150 or so Mb usable memory.

[carpenike]

Excellent, that fixed it!

Thanks!

Code: Select all

homefileserv ~ # cat /proc/meminfo
MemTotal:      1036564 kB
MemFree:        865200 kB
Buffers:         21236 kB
Cached:          42736 kB
SwapCached:          0 kB
Active:         119840 kB
Inactive:        38176 kB
HighTotal:      131008 kB
HighFree:          256 kB
LowTotal:       905556 kB
LowFree:        864944 kB
SwapTotal:      987988 kB
SwapFree:       987988 kB
Dirty:              36 kB
Writeback:           0 kB
Mapped:         105328 kB
Slab:             7832 kB
CommitLimit:   1506268 kB
Committed_AS:   304852 kB
PageTables:        744 kB
VmallocTotal:   114680 kB
VmallocUsed:      3440 kB
VmallocChunk:   111240 kB

[ts77]

any known problems with net-snmp 5.4?

out of the blue after upgrading net-snmp to 5.4 (on the cacti-box) I got all memory graphs at 100% real memory used and when trying the script I get output like that:

Code: Select all

 php ./netsnmp_memory_usage.php tcp:127.0.0.1, public, 1, , , 16109, 500
totalReal:3082408 availReal:0 totalSwap:0 availSwap:0 memBuffer:0 memCached:0 usedReal:3082408 usedSwap:0

Any ideas what went wrong here and how to solve this?

thanks,

thomas

[ehall]

[post deleted--it was in the wrong thread entirely]

[ts77]

Hmm, I recompiled net-snmp now with lm_sensors but strange thing is, I didn't even have lm_sensors installed before.
Is that really required on the cacti-machine or more on the remote machines which are monitored.

restarted net-snmp on the cacti-machine now but I still get

Code: Select all

snmpwalk -v 2c -c public 192.168.115.11 .1.3.6.1.4.1.2021.13.16
LM-SENSORS-MIB::lmSensors = No Such Object available on this agent at this OID

guess I have to read up more how to enable lm_sensors if thats really what is needed here.

[ehall]

[post deleted--it was in the wrong thread entirely]

[davibou]

I have a problem whith this script ...

When i do
php -q netsnmp_memory_usage.php 127.0.0.1, public, 1, , , 161, 500

it give me that

totalReal:254848 availReal:19128 totalSwap:524280 availSwap:483900 memBuffer:10124 memCached:50516 usedReal:175080 usedSwap:40380


But the pooler write warning in the log , grph is showing nothing

Code: Select all

02/05/2007 08:35:02 PM - CMDPHP: Poller[0] Host[1] DS[3] CMD: perl /opt/cacti/scripts/linux_memory.pl MemFree:, output: 13128

 02/05/2007 08:35:04 PM - CMDPHP: Poller[0] Host[1] DS[107] CMD: php /opt/cacti/scripts/netsnmp_memory_usage.php 127.0.0.1, public, 1, , , 161, 500, output: U

02/05/2007 08:35:04 PM - CMDPHP: Poller[0] Host[1] DS[107] WARNING: Result from CMD not valid. Partial Result: Content-type: text/html
X-Powered-By: PHP/4.3.9

What could i do ??

[exploit-soget]

I have exactly the same problem.
When I do this command :

/usr/local/bin/php netsnmp_memory_usage.php superexpl, public, 1, , , 161, 500
totalReal:2052988 availReal:43952 totalSwap:4200956 availSwap:4200944 memBuffer:317168 memCached:1017712 usedReal:674156 usedSwap:12

BUT NOTHING IS GRAPH AND RRDTOOL SAYS OK IN THE DEBUG MODE

[ehall]

The posted update contains a fix that converts the SNMP timeout value from Cacti's millisecond format to PHP/Net-SNMP microsecond format. The lack of conversion in the earlier script would sometimes result in excessive retries or complete failure.

[Yzo]

I'm also having the same problem that ts77 reported in his post of 1/31/07.

The PHP script is no longer reporting the proper values for availReal, totalSwap, availSwap, memBuffer, memCached, usedReal, and usedSwap. The only value that is accurate is totalReal.

Example:

Code: Select all

$ php netsnmp_memory_usage.php 127.0.0.1, public, 2, , , 161, 500
totalReal:2067456 availReal:0 totalSwap:0 availSwap:0 memBuffer:0 memCached:0 usedReal:2067456 usedSwap:0

If I manually walk the OIDs the script is checking for, I do get the proper values returned.

Examples:

Code: Select all

$ snmpwalk -v 2c -c public 127.0.0.1 .1.3.6.1.4.1.2021.4.15.0
UCD-SNMP-MIB::memCached.0 = INTEGER: 1427728 kB

$ snmpwalk -v 2c -c public 127.0.0.1 .1.3.6.1.4.1.2021.4.14.0
UCD-SNMP-MIB::memBuffer.0 = INTEGER: 59492 kB

$ snmpwalk -v 2c -c public 127.0.0.1 .1.3.6.1.4.1.2021.4.6.0
UCD-SNMP-MIB::memAvailReal.0 = INTEGER: 114640 kB

Here the output of free:

Code: Select all

$ free
             total       used       free     shared    buffers     cached
Mem:       2067456    1984120      83336          0      61224    1448452
-/+ buffers/cache:     474444    1593012
Swap:      1959920        220    1959700

Like ts77, this seems to have started immediately after an update to Net-SNMP 5.4. Downloading the updated script has not helped.

....

Actually, after looking over the script one more time and looking at what I pasted into this post, it looks like the Net-SNMP update has now caused all the OIDs to report "kB" after every value instead of just after totalReal.

After commenting out the IF/ELSE statement in the script so that all the OIDs get parsed, the output seems to be back to normal.

Code: Select all

$ php netsnmp_memory_usage.php 127.0.0.1, public, 2, , , 161, 500
totalReal:2067456 availReal:72428 totalSwap:1959920 availSwap:1959700 memBuffer:62976 memCached:1471788 usedReal:460264 usedSwap:220

Not sure if this calls for a script update, or if I should just find a way to get snmp not to but kBs at the ends of those values.

[Suhy]

I haven't been able to get this report to work, or any other report that involves load average or memory stats. I just don't get any data back from certain MIBs. (ex: .1.3.6.1.4.1.2021.4.6.0)

An snmpwalk shows results such as:

UCD-SNMP-MIB::memAvailReal.0 = No more variables left in this MIB View (It is past the end of the MIB tree)

Other stats and graphs work fine (ex: ethernet traffic) so I know I have snmp and cacti setup correctly.

Is there something I need to do/edit/upgrade on my servers to get these other stats?

I have similar problem

I don't get anything from graph..
if i run php poller manualy i get :: ERROR: Not enough arguments
Im using centos 4.4 and net-snmp 5.1.2-11el4.7..also i have cacti 0.8.6j

how do i alter snmp cpnfiguration to allow viewing of mib data other than [system]?

[pheldens]

change "netsnmp_memory_usage.php" from line 176 to the following to fix "0" output in newer net-snmp.

Code: Select all

        #
        # make sure the data only contains numbers
        #
        if (is_numeric($cmdoutput[$cmdoutputcount]) == FALSE) {

        //      #
        //      # the totalReal OID appends "KBytes" to the value, so eliminate it
        //      #
        //      if ($oid[$cmdoutputcount][0] == "totalReal") {

                        $cmdoutput[$cmdoutputcount] = intval($cmdoutput[$cmdoutputcount]);
        //      }

        //      else {
        //              $cmdoutput[$cmdoutputcount] = 0;
        //      }
        }

[ehall]

I uploaded a new tar.gz which contains an updated script that should fix the problem. If the value is not numeric, it checks to see if the trailing characters are "kB" or "KBytes".

[Suhy]

Problem still remains.

i can import both templates, and i copy php file, after that i chmod to cactiuser on php file, but graph is empty, and php poler.php shows: ERROR: Not enough arguments

i hope this will be implemented in next cacti version

[sgn008]

Hi,

Does anyone having problem when graph memory on SuSe Linux 64GB ?

Apparently my graph shows that all 64GB is shown as Real Usage. There is no "Buffer", "Cache", "Free". This contradicts with my top command.

By the way, I use this template with RedHat ES 16GB with no problem.

Any suggestion is greatly appreciated.