sh.exe, php.exe, spine.exe problems

[africanw]

Hi,

Recently moved our cacti install from a Windows 2003 32bit machine to a bigger Windows 2008 R2 64 bit machine. Am running the latest cacti and spine with php 5.3.6. You can see the cpu spikes below with them dropping when the script runs.

I have the latest cygwin install and all the paths have been added but am having a lot of problems with sh.exe, php.exe, spine.exe not exiting correctly and after a few hours they just use all the cpus (2 x 4c).

Currently have a script running that kills them off every two hours.

D:\scripts\pskill.exe spine.exe
D:\scripts\pskill.exe php.exe
D:\scripts\pskill.exe sh.exe
D:\scripts\pskill.exe cmd.exe

Any help would be great.

Thanks

Info:

General Information
Date Mon, 16 May 2011 08:49:59 +1000
Cacti Version 0.8.7g
Cacti OS win32
SNMP Version NET-SNMP version: 5.5
RRDTool Version RRDTool 1.2.x
Hosts 39
Graphs 506
Data Sources Script/Command: 30
SNMP: 274
SNMP Query: 278
Script - Script Server (PHP): 51
Script Query - Script Server: 177
Total: 810
Poller Information
Interval 60
Type spine
Items Action[0]: 1034
Action[1]: 30
Action[2]: 273
Total: 1337
Concurrent Processes 4
Max Threads 8
PHP Servers 6
Script Timeout 50
Max OID 10
Last Run Statistics Time:25.1806 Method:spine Processes:4 Threads:8 Hosts:40 HostsPerProcess:10 DataSources:1058 RRDsProcessed:564
PHP Information
PHP Version 5.3.6
PHP OS WINNT
PHP uname Windows NT CANBERRA 6.1 build 7601 (Unknow Windows version Enterprise Edition Service Pack 1) i586
PHP SNMP Installed
max_execution_time 300
memory_limit 128M

[BSOD2600]

typically this is caused by rouge scripts which do not properly self terminate. Use a tool like sysinternals process explorer to find out what script(s) (via the command line parameters) are running for those hung processes.

[africanw]

Thats it was two of the standard perl ones getting stuck (linix mem,logged in users).

[BSOD2600]

you're running those scripts against your windows pc?

[africanw]

No against target linux machines. We have Windows and linux machines in our network and the cacti machine is Widows as well. It seems to be a Perl problem and does not matter if I use cygwin perl or active perl but sh.exe hangs on the perl scripts very randomly.

Example of one getting stuck:
D:\Apps\cygwin\bin\sh.exe -c "D:\\Apps\\Perl\\bin\\perl.exe D:/_Webroot/stats.ecorner.com/scripts/loadavg_multi.pl"

[BSOD2600]

Gotcha.

Yea, these scripts really should be improved to include some sort of timeout/self termination logic.