net.if.in and net.if.out - bad values, showing GBps?

[stever]

I'm monitoring some hosts with net.if.in (using zabbix_agentd), i.e. net.if.in[hme0], and its throwing back values around 10GBps. Kind of hard to do with a 100Mb network card.

Has anyone had this happen? Is there some scaling I should be doing? I'm getting the same results on both my linux and solaris systems I'm monitoring.

Also - I do have this setup as a differential value, calculating / second, and not an absolute value.

[ypircc]

Under the items configuration you need to change "Store value" to "Delta speed per second" this should solve your problem.

[stever]

Under the items configuration you need to change "Store value" to "Delta speed per second" this should solve your problem.

Hi - I did do that - I couldn't remember what the name of the option was, so in my post i listed it as per second. I even double-checked the DB, and the values coming in (set with delta speed per second) were 10+ Gbps...

[achu_sab]

For getting correct value try this:
Units - b/s
Use multiplier -Custom Multiplier
Custom multiplier - 8
It will show the bandwidth usage in bits / sec . Try this

[stever]

Hi - I have it configured like this already as well.

The problem is that the values are just not correct. I have some other monitoring software going, and when in reality I am pushing around 0 - 10 kbps, zabbix is reporting values of around 10,000,000,000 . And yes, I am already using the "Delta speed per second" setting - zabbix is telling me I am getting around 10Gbps.

[sauron]

2stever: clear history. Zabbix don't convert previos data to new multiplyer.

[fips]

I had the same problem with AIX and I figgered out it was a problem with 64 bit counters on the AIX. I set up some UserParameters which dived the counters befor sending to the zabbix server.

[stever]

I tried recompiling the agent as a 64 bit binary (-m64 flag to gcc), and still no luck.

What were the values you divded by to get something meaningful?

[fips]

In AIX I have set up the UserParameter as following

UserParameter=usr.net.in[*],netstat -v $1 | grep Bytes | awk '{printf("%d", $ 4%2^24)}'
UserParameter=usr.net.out[*],netstat -v $1 | grep Bytes | awk '{printf("%d", $ 2%2^24)}'

[stever]

Thanks for those - I didn't have to do any division on mine, just setup a similar UserParameter in the config files - thanks for the help.

A little ugly and probably can be done better, but:

UserParameter=usr.net.in[*],netstat -v -I $1 | grep -v Name | head -1 | awk "{print $ 5}"
UserParameter=usr.net.out[*],netstat -v -I $1 | grep -v Name | head -1 | awk "{print $ 7}"

worked for me. This is on Sun sparc systems, running both Solaris 8 and 9 - haven't tried it on anything else, but will probably work across most solaris installs.

[themons]

I have the same probleme but only with the netloadin: its in Gbs.
Is it a probleme of a 64bit number value. When i use paquets no porblems.

[themons]

a new solution

An ather loution is to check every 30 seconds the number or packets

like this :

Key : net.if.in[hme0,packets]*1500/30 (1500=mtu)(30=update interval)
Units : b/s
Update interval : 30
Sore value : Delta (Simple change)