GRAPH shows nothing below 10hour scale

[Ralf Steiner]

When we look at the graph of cpu load 24h scale we see a perfect graph.
But when we scale down to 14 hours suddenly there are gaps in the value lines. In the 13 hour scale view there is only left one little value line on the left and one in the middle.

Every time schedule below 12 hours shows an empty graph without values.

This problem can be seen in 1.4.1 and (pre)1.4.2.

Any hints on this phenomena?

Kind regards,
Ralf Steiner

[Aly]

Check 'Update interval' of that items..

[Ralf Steiner]

The update interval is 5 seconds.
This is the default value. Should
it be higher or lower?

[Aly]

No, it's ok. Hmm maybe there is no data on that period.. So when you take bigger period it's connects the whole graph, but when you look closer(smaller period) you see gaps.

Could it be missing data in db on that period?

[Ralf Steiner]

the "500 latest values" of processor load shows this:

2007.Jul.19 00:36:35 0.6500
2007.Jul.19 00:22:09 0.0800
2007.Jul.19 00:07:44 0.7800
2007.Jul.18 23:53:18 0.5000
2007.Jul.18 23:38:56 0.1200
2007.Jul.18 23:24:35 0.6000
2007.Jul.18 23:10:20 1.5700
2007.Jul.18 22:56:01 2.1800
2007.Jul.18 22:41:18 1.5300
2007.Jul.18 22:26:27 0.4900

so the cpu values are not recorded
every 5 seconds but 15 minutes in
the database !?

[Alexei]

It seems you have huge performance problems.

[Ralf Steiner]

I agree that the server load is more than
it is on an idle server.
Sometimes we really have heavy load on
the system. Then CPU load rises up to
4.0 or 6.0 for some time. This is why we want
to monitor it with Zabbix .

But during the snapshot times i posted
the server load is between 1.0 and 2.0.

This is not a huge performance problem
for the server. The application performance
during this situation is perfect.

Does Zabbix stop working when CPU values rise
over 1.0 ?

[Ralf Steiner]

Here is a hardcopy of another value: CPU Idle Timer.
I shows that there are gaps even at 04:18 o'clock
in the morning when the server is really idle.

When i reduce scaling to 10 hours there is no green
graph any longer.

[Ralf Steiner]

Help, please!

[Aly]

Definitely the problem not in the graphs, there is to few data in your DB to draw it correctly. It seems that generally the problem is that data isn't coming every 5 sec, as it should be..

[Ralf Steiner]

We got it!

The SNMP system simply did not work. The
agent asked 127.0.0.1 and the snmpd listened
on the eth ip address.

The waiting delays caused the other agent
activities to time out. So after this fix we get
pretty much values. Every five seconds

Thank you for your patience.