proc.num always shows 1 running procs when its zero

[worksyn]

Hi,

I am using proc.num[,,run] as monitoring item (Zabbix agent type) on my linux server (CentOS 6.x), however even if no running processes on the server - it always shows - 1.

the same result if i use zabbix_get:

[root@zabbix ~]# zabbix_get -s 10.39.81.57 -p 10050 -k proc.num[,,run]
1

at the same time on the monitoring server:

[root@rets ~]# vmstat 1
procs -----------memory---------- ---swap-- -----io---- --system-- -----cpu-----
r b swpd free buff cache si so bi bo in cs us sy id wa st
0 0 0 5642460 0 2656752 0 0 2 1 0 15 1 0 99 0 0
0 0 0 5642460 0 2656744 0 0 0 12 0 184 0 0 100 0 0
0 0 0 5642460 0 2656744 0 0 0 0 0 170 0 0 100 0 0
0 0 0 5642460 0 2656744 0 0 0 0 0 317 0 0 100 0 0

I guess, zabbix agent for finding number of running procs looks into /proc/loadavg file:

# cat /proc/loadavg
0.00 0.00 0.00 1/2277 20091

it always shows 1 - active task in 4th column.

it shows correct result if I use user name:

[root@zabbix ~]# zabbix_get -s 10.39.81.57 -p 10050 -k proc.num[,root,run]
0

but for all users it shows - 1.

is it bug?


ps. my zabbix server and agent version: 2.4

zabbix-server-2.4.5-1.el6.x86_64
zabbix-agent-2.4.5-1.el6.x86_64

[LenR]

Maybe the agent making the check is counted as a running process.

[worksyn]

Maybe the agent making the check is counted as a running process.

anyway it's not accurate, and should be fixed

[Atsushi]

I think correct that a process in running is zabbix agent.

zabbix_agentd is run by zabbix user.
Please check proc.num[,zabbix,run] value.

Therefore, I think that it is functioning normally.

[worksyn]

I think correct that a process in running is zabbix agent.

zabbix_agentd is run by zabbix user.
Please check proc.num[,zabbix,run] value.

Therefore, I think that it is functioning normally.

well, u are right:

[root@zabbix ~]# zabbix_get -s 10.39.81.57 -p 10050 -k proc.num[,zabbix,run]
1

but that process runs for very short period of time, i suppose its 0.05 second or similar. But on the graph I see that system always have 1 process in running state - which obviously not correct.

[worksyn]

# ps -eo pid,user,stime,cputime,stat,comm | sed -n '1p;/zabbix/p'
PID USER STIME TIME STAT COMMAND
1201 zabbix Jun03 00:00:00 S zabbix_agentd
1204 zabbix Jun03 00:00:16 S zabbix_agentd
1205 zabbix Jun03 00:00:08 S zabbix_agentd
1206 zabbix Jun03 00:00:08 S zabbix_agentd
1207 zabbix Jun03 00:00:08 S zabbix_agentd
1208 zabbix Jun03 00:00:02 S zabbix_agentd

for the last day zabbix processes have been running for 42 seconds, it is 42/(3600*24) = 0.0004 processes per day. I suppose such kind of check should exclude zabbix agent process

[Atsushi]

It will always be run state at the moment of getting value.
I think the value is not in the less than 1.

It can not be confirmed at the ps command
because it is the change in instantaneous state.

[worksyn]

It can not be confirmed at the ps command
because it is the change in instantaneous state.

I understand that, but i am talking about Graph, that shows that my server always have 1 running process, which is not true:

[worksyn]

guys, I want that you pay attention to this, because your competitor ganglia doesn't have such issue. In Running processes graph it doesn't calculate gmond process(ganglia agent)