Graph with large spike/peak.

[robertdb]

Hi,

We have a few graphs that show a spike or peak from time to time. I guess this user experienced the same.

My analysis is that Zabbix was not able to read data for some time, after which it read a much larger value of the device.

Here is some information about the analysis.

The spike in the graph:


Values from the database:

mysql> select * from history_uint where itemid=46069 and clock > 1336039154 and clock < 1336042754;
+--------+------------+-------------+
| itemid | clock      | value       |
+--------+------------+-------------+
|  46069 | 1336039218 |  4341792000 |
|  46069 | 1336039277 |  2724783880 |
|  46069 | 1336039334 |   644267520 |
|  46069 | 1336042515 | 88831519920 |
|  46069 | 1336042577 |  1286807600 |
|  46069 | 1336042635 |   154926200 |
|  46069 | 1336042696 |   244149880 |
+--------+------------+-------------+
7 rows in set (0.02 sec)

The fourth value (88831519920) is much higher than the others, if you take a look at the clock value, you will notice there is a gap of approximately 3000 seconds or so.

Some information about this item:
Type of information: Numeric (unsigned)
Data type: Decimal
Store value: Delta (Simple change)

In order to get the graph right, I guess there are several solutions:

1. Zabbix should not store the first value after a period of no-data.
2. Zabbix should not graph the first value after a period of no-data.
3. Zabbix should graph the average over the time of the no-data period.

Does anybody recognise this behavior or have a solution?

Regards, Robert de Bock.

[eskytthe]

Your root problem is that zabbix server/agent does not get or are not able to get data for long time. You have to dig into that.
Maybe the host your are monitor have real performance problems - e.g. look at other items for that host in same time spans.
Br
Erik

[robertdb]

Your root problem is that zabbix server/agent does not get or are not able to get data for long time. You have to dig into that.
Maybe the host your are monitor have real performance problems - e.g. look at other items for that host in same time spans.
Br
Erik

I agree that it's a problem that this box was not responding, but there are different situations where the graphing could go wrong:

1. A maintenance period without capturing data
2. A planned reboot/upgrade
3. Maintenance for the Zabbix server or database

So, I still think the algorithm Zabbix uses to graph stuff is incorrect.

[eskytthe]

This may be related even that it is data/sec:

https://support.zabbix.com/browse/ZBX-4884

(fixed in 1.8.13)
/Erik

[robertdb]

This may be related even that it is data/sec:

https://support.zabbix.com/browse/ZBX-4884

(fixed in 1.8.13)
/Erik

I think you are right, that might well be the same problem. The solution would be to update to 1.8.13, which is stable. I'll try, thanks a lot for that link!