Ad Widget

**richlv** · 20-01-2010, 21:30

it's a bit hard to look at the misaligned output (maybe try formatting it in [code] block), but i couldn't see anything "really high" in there. which value exactly caught your attention ?

**iskondi** · 20-01-2010, 21:35

Sorry, you're right that is hard to read.
What I'm seeing is that cumulative CPU time on the zabbix_agentd process is near the top on both boxes.

It's not that the boxes themselves are over loaded but that the zabbix_agentd, even on a machine that isn't being monitored is accumulating a lot time.

On the non-monitored machine it's up to 8 hours of time and on the monitored machine it's up to 20 hours.

Let me figure out how to format the TOP output better but the gist of it is why is an unmonitored or monitored machine for that matter accumulating so much CPU time?

Thanks!
d

**richlv** · 29-01-2010, 20:54

i haven't paid much attention to cumulative times agent daemon achieves - maybe the machines are so underloaded that agent stands out because of active item scheduling and such ?
unless you see it in current load lists as the bigger load generator, i'd personally not worry much about that...

**henry** · 24-06-2010, 07:02

I noticed the same issue here on a stock Solaris10 zabbix_agentd (in a S10 zone).

The box is idle at the moment and I even turned off almost all the alarming in the template but the the load still pops up from 0.02 to a 0.2!

Also the CPU time that the agent chews up is about 120 minutes in a day or two!

Compare that to the openfire server that's running on the same machine with a few people chatting away all day and it only accumulates 12 minutes in 5 days!

Guess someone who understands the source code for the Solaris agent should have a look at the code and see what it's doing?

regards, henry

**jerrylenk** · 28-06-2010, 11:56

Hi henry,
turning on or off some alarms cannot have any effect on the agent side, as this is done on the server.

Hi iskondi,
coul it be that the cumulated time in top is not hours but minutes:secs.mils ?

Since I rebooted one of my servers ~1 hour ago, I had a look at it:
There are 6 agentd processes running, 3 of which are among the top ten consumers of CPU accumulated time --
but since reboot and start of the agentd, CPU idle time has been between 99% and 100% all along, so for me there is nothing to worry about.

Regards, Jerry

**henry** · 11-07-2010, 10:01

[QUOTE=henry;67990]

Originally posted by jerrylenk

Hi henry,
turning on or off some alarms cannot have any effect on the agent side, as this is done on the server.
.

Good point...

well, I tried to cut down the checks to less than three, but the CPU consumption is still high:

zabbix 14371 14370 1 Jun 24 ? 4055:37 zabbix_agentd

On S10 SPARC local zone..

**trikke** · 29-09-2010, 13:55

Hi,

any solution to this Problem, as I'm having the same issue?

Greets
Patrick

**henry** · 11-10-2010, 12:01

Originally posted by trikke

Hi,

any solution to this Problem, as I'm having the same issue?

Greets
Patrick

Unfortunately not. It might need a programmer with Solaris and dtrace experience to track down where the agent chews up all that CPU.

Here, it racked up more than 2000 CPU minutes in a couple of months! And that doesn't seem to change if the zabbix server is not even polling the agent. Keeps the load at a minimum of 0.2, whereas before it was 0.0.

Cheers,
heinz

Ad Widget

Is zabbix_agentd using a lot of CPU?

Is zabbix_agentd using a lot of CPU?

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment