Delta (Speed per second) strangeness

[novazion]

Hello,

I'm using for monitoring network interfaces net.if.in and net.if.out
I use them trough some network discovery.

If use the default template refresh item rate (every 60 seconds) values are ok.

But if I change the refresh item rate to 3600 seconds (in my case) values doesn't match the reality anymore:

2014.Oct.08 11:15:03	66728360
2014.Oct.08 11:14:03	65780936
2014.Oct.08 11:13:03	65294848
2014.Oct.08 11:12:02	64569048
2014.Oct.08 11:11:02	64452928
2014.Oct.08 11:09:02	66195128
2014.Oct.08 11:06:52	798552
2014.Oct.08 10:06:49	1319048
2014.Oct.08 07:06:39	6295800
2014.Oct.08 05:06:34	2182560
2014.Oct.08 04:06:30	6518440

My server version is : Zabbix 2.2.6

Am I missing something?

Thanks.
Emeric.

[novazion]

Is this normal, is this a bug.

Does anyone has the same issue?

Should I report a bug?

Thanks.
Emeric.

[kloczek]

Hello,
[..]
But if I change the refresh item rate to 3600 seconds (in my case) values doesn't match the reality anymore:

2014.Oct.08 11:15:03	66728360
2014.Oct.08 11:14:03	65780936
2014.Oct.08 11:13:03	65294848
2014.Oct.08 11:12:02	64569048
2014.Oct.08 11:11:02	64452928
2014.Oct.08 11:09:02	66195128
2014.Oct.08 11:06:52	798552
2014.Oct.08 10:06:49	1319048
2014.Oct.08 07:06:39	6295800
2014.Oct.08 05:06:34	2182560
2014.Oct.08 04:06:30	6518440

My server version is : Zabbix 2.2.6

Am I missing something?

Thanks.
Emeric.

Lets calculate ..

$ bc
2^32
4294967296
2^64
18446744073709551616
66728360*3600
240222096000

So 66728360*3600 is greater than 2^32 and your hourly calculated delta per second values looks quite random.
Looking on https://www.zabbix.com/documentation...fig/items/item "numeric (unsigned)" type it should be 64bit unsigned int so it should be OK.
It is possible that during calculate delta per second somewhere is used only 32bit data type.
I think that you should try to open jira ticket on http://support.zabbix.com/

[novazion]

Thanks a lot.
I would never be able to point out this scenario.

Nice one.

So i'm going to try the ticket and I'll try to find a workaround in the same time.

[aib]

btw, SNMP has two different branches to collect data - 32 bit and 64 bit.
maybe, you have the same issue with overflowing because you try to collect data from wrong interface branch?

http://www.cisco.com/c/en/us/support/docs/ip/simple-network-management-protocol-snmp/26007-faq-snmpcounter.html