Win32 - Monitor running processes

[b3nw]

I have 6-7 servers with graphs created. I just added the processor page, and when I try to select any of the processes it lists, it throws the error. But for example, interface statistics, mounted partitions, all map fine. I've not had any issues until I tried to add a graph for this script/template

as for versions, standard debian 5.0 stable, i'll have to get the exact versions first thing tomorrow morning.

[BSOD2600]

Hmm, odd. The devices generating those errors are Windows boxes?

[b3nw]

yes, windows 2000 server.

[BSOD2600]

The verbose query and/or manual snmpwalk of .1.3.6.1.2.1.25.4.2.1 returns correct data for that problem device?

[intenso]

hi,

my problem:

if i close the monitored program (like iexplore.exe) and restart it the mem/cpu line on the graph stops forever.

in the data source i have changed the "index type" from "hrSWRunIndex" to "hrSWRunName". on the "index value" field i removed the pid and write the proces name.
but cacti ignored it. in the cacti log file the oid for this proces will be never changed or updated to the new pid.

cacti Log:
09/16/2009 02:50:25 PM - SPINE: Poller[0] Host[126] DS[2214] SNMP: v1: 192.168.10.35, dsname: proc_mem, oid: .1.3.6.1.2.1.25.5.1.1.2.1336, value: U

(1336 is the old PID)


can anyone help me please?

thanks

[BSOD2600]

Until cacti runs another reindex for the 'running process info' query, it'll still have the old/invalid PID of the process. You have two options:

1) leave the Reindex method for 'SNMP - Running Processes' as 'Index Count Changed' -- then whenever the number of processes changes from its current number, cacti will invoke a reindex and the graph will work again.
2) Change the reindex method to Verify all fields.

[intenso]

hi,

thanks for your answer and sorry for my english...

1. the "Index Count Changed" method is my actual method.

2. i have changed the "re-indexed method" to "verify all fields".
if i stop and start my proces the pid is still the old in cacti.

then if i go to the device and klick on the "verbose query" button, cacti re-indexed the pid-map and i see the correct pid for my proces.

but this is an unusfull "handjob" to do this on every time if a proces restarts...

how reindexed cacti automatic? why doesnt work it?

thanks

[BSOD2600]

Not sure what you've done to mess up the script for your install... I know that for me, if I close the process being monitored (say trillian.exe) the current process count drops from 41 to 40. If cacti polls the server during that time, it'll notice the different in processes and issue a reindex. Now, if during that same 5 minute time window, I launch another process so the process count is STILL 41, cacti will NOT issue a reindex, even though there are now different processes/PIDs. As stated earlier, the only way around this is to use the Verify all fields option.

Look through your cacti log file while in debug mode during a polling cycle. Find the Data Source (DS ##) for this query and track what cacti does. Why does/doesn't it do a reindex?

[intenso]

hi,

ok, i have changed the cacti default "reindex method for Data Queries" to "Verify All Fields" and now it works

thanks for help.

[pirx]

nice template and it even works with net-snmp linux clients. At least on my debian lenny systems with net-snmp 5.4.1.

[jpipitone]

I am using this great template to keep an eye on w3wp.exe memory usage.

We have multiple application pools which each spawn their own instance of w3wp.exe. When I add all of these processes, they each have w3wp.exe as the label.

Is there a way to name these graphs individually so I know which application pool (Pool_sitename) is associated with each w3wp.exe process?

[BSOD2600]

SNMP doesn't expose this info. You'll need a WMI script instead to handle the app pools

[adrianzzs]

hello BSOD2600
your script is very useful !

still, there is any way to have the datasources automatically configured by hrSWRunName with the correct index value, for example explorer.exe instead of hrSWRunIndex with the PID ?

I want this because I want to monitor a lot of services but when they restart, a new PID is generated and the only way that re-index works is if I have hrSWRunName with the correct service name in Index Value field.

thank you

[BSOD2600]

It should be using the process name by default...

[adrianzzs]

hello
Thank you for your answer, but I don't think you understood me

so I have my host, and I click on create graphs for this host


I also have the data queries for my host "SNMP Running Processes" with Index count changed for reindex method:


Then I will select the process that I want to monitor (explorer.exe) and hit the create button


Next I'll have the possibility to give the name to my graph, by default it's Hostname Process + hrSWRunName (I'm not referring to this in my previous post) I dont have a problem with the name of the graph


Now If I go to my device and select the datasource that was just created


I'll see the following screen:


Now here I have a problem ! I want this to be automatically with
hrSWRunName in the Index Type field and not hrSWRunIndex
and
the correct process name in the Index Value field and not the PID value.

Do you think it's possible to have that ?
Thanks