proc.num Bug?

[tighep]

I've got an issue with proc.num, and I think it might be a bug. Running Zabbix 1.4.4, RedHat EL 4 is the agent side. I've got 3 processes that I want to watch.

cbpadmin 5393 1 0 Feb13 ? 00:00:00 cbp_move_request -d
cbpadmin 5395 1 0 Feb13 ? 00:00:00 cbp_move_reply -d
cbpadmin 5397 1 0 Feb13 ? 00:03:27 cbp_handle_event -d

proc.num[cbp_move_reply] returns 1 as expected, but
proc.num[cbp_handle_event] and proc.num[cbp_move_request] both return 0, could this be an issue with the string length being a full 16 characters? I've tried monitoring other processes, and have seen similar behavior especially when trying to watch a process that is fully qualified.

[tighep]

Here's confirmation of my condition

[root@mqdev sbin]# ./zabbix_agentd -t proc.num[cbp_move_request]
proc.num[cbp_move_request] [u|0]
[root@mqdev sbin]# ./zabbix_agentd -t proc.num[cbp_move_reques]
proc.num[cbp_move_reques] [u|1]
[root@mqdev sbin]# ps -ef |grep cbp_move_request
cbpadmin 5393 1 0 Feb13 ? 00:00:00 cbp_move_request -d

[skogan]

I had similair problems - I think it's some kind of out of bounds issue or something.

My solution is to use the command line matching instead. Like this:

proc.num[,,,cbp_move_request]

[tahir.bhatti]

I am also facing same problem..., I think its a bug in 1.4.x. I simply tried to count the No of MS Word Runnin Precesses, there are multiple processes are running on the machine but in monitoring it showed just one.

[skogan]

Are you sure those winword processes are not actually threads? Threads are not supposed to be counted.

[tighep]

I had similair problems - I think it's some kind of out of bounds issue or something.

My solution is to use the command line matching instead. Like this:

proc.num[,,,cbp_move_request]

This is how I've done it for other things as well, but there is one flaw. I've got a process that gets started with fully qualified path, that's fine. But if it has to restart, the stupid app restarts with ./<appname>, which now causes my trigger to alarm. I believe this doesn't trip if you use proc.num[<appname>], but it's a long app, and won't match due to the size limit.

[skogan]

You don't necessarily have to use the full path in the "cmdline" option. You can use any part of it and it will just match it. That's what I do, anyhow.