Monitoring Bandwidth

[IanStewart_Inoapps]

All,

I'm looking to find out whether Zabbix is capable of monitoring bandwidth usage.

We are a hosting company, and, have a remote client, who utilises SSL VPN to remotely log onto their hosted application.

To ensure that we provide a suitable level of service, I'm keen to find out whether Zabbix has a method of monitoring the firewall, or switches to find out if bandwidth drops.

IF it can be done, what specifically would I be monitoring, bandwidth, or can it be as sophisticated as monitoring actual MB/GB download speeds?

[slavik]

you can monitor speed from firewall via snmp.
use my utility to create host xml from snmp enabled router/firewall/switch (http://www.zabbix.com/forum/showthread.php?t=20856)
after that you must import result in xml form into zabbix, and create speed drop triggers.

[untergeek]

Yes, bandwidth monitoring is there. It can be as simple or sophisticated as you want it to be. The default templates tend to sample every 5 seconds for some network data, you'll probably want to tune that.

[IanStewart_Inoapps]

How do I?

Yes, bandwidth monitoring is there. It can be as simple or sophisticated as you want it to be. The default templates tend to sample every 5 seconds for some network data, you'll probably want to tune that.

which template and item/trigger would I use then for monitoring bandwidth?

I can find out the firewall details, but, unsure which Zabbix template to apply.

HELP!!!!!

[untergeek]

Some questions first:

Is your firewall a server (e.g. a Linux box running ipfw or iptables or something else) or a hardware device from a company like Cisco where you'd query the device via SNMP?

You may have to make your own item/trigger to monitor. The instructions are not necessarily the easiest to follow in the online manual, but they will make sense once you've made one or two items. We're usually happy to help here, but are busy ourselves and not likely to just try to make your configuration for you.

Basically, you need to first be collecting data. If it's a linux box, you'll be monitoring in/out-bound traffic on eth0 or eth1 or something. It'd be an SNMP OID for a device. You'll make a trigger based on a value you wish to be informed of, e.g. if traffic exceeds 500Mbits/sec.

Good luck, keep us posted if you need more assistance.

[lupick]

Smokeping?

Dear all,

is it possible to monitor ping latency like smokeping? I've used cacti and it has some nice graph based on fping result.

I don't have access to firewall/router cause they are managed by external supplier.

Thxs

Luca

[untergeek]

Try a new item with type Simple Check, the key should be:

icmppingsec

Which will return the ping time in seconds (be sure to label units as s) or a straight up 0 in event of a ping failure.

[lupick]

Try a new item with type Simple Check, the key should be:

icmppingsec

Which will return the ping time in seconds (be sure to label units as s) or a straight up 0 in event of a ping failure.

In this case I'll have the ping reply time. is it possible to have on the same graph also the ping latency and ping loss like :

http://oss.oetiker.ch/smokeping-demo/?target=Customers.OP.johan

If this is not possible can I add an HTML frame to Zabbik in order to link the smokeping site inside zabbix?

Thank you

L.

[untergeek]

I'm confused. What's ping latency if not the delay in return time?

In any case, ping loss also has a Simple Check: icmppingloss[<ip>,<count>,<interval>,<size>,<timeout>]

You could make a graph with both icmppingsec & icmppingloss and have those things.

[lupick]

with somkeping in the same graph you can read:

the averange of the ping latency
the variation of the different ping lataency.
the packet loss

http://oss.oetiker.ch/smokeping/doc/reading.en.html

BR

L.

[untergeek]

You're only serving to reinforce my previous argument. The ping "time" is the ping "latency." In the link you provide it's referred to as "rtt" for round-trip time. They are the same thing. With the item provided (icmppingtime) zabbix will trend them over time. With a graph for said item you'll see the average as well as highs and lows (variation). If you add packet loss to the graph (using icmppingloss) you could generate a similar graph. It won't be the same, but you could do something like it using the "bar graph" type. The one thing Zabbix should not be able to do, graph-wise, is the color coding of different levels of ping loss. Other than that, you should be able to match fairly closely.

icmppingloss[<ip>,<count>,<interval>,<size>,<timeout>]

The link you provided says it was 20 pings every 300 seconds. Try:

icmppingloss[,20,,,]
icmppingtime[,20,,,] will do the same for icmppingtime, but a single ping is possible without the bracketed part (the remaining items will all be the default values for zabbix, which you can research on your own).

Set the time between checks to be 300 seconds.

[lupick]

You're only serving to reinforce my previous argument. The ping "time" is the ping "latency." In the link you provide it's referred to as "rtt" for round-trip time. They are the same thing. With the item provided (icmppingtime) zabbix will trend them over time. With a graph for said item you'll see the average as well as highs and lows (variation). If you add packet loss to the graph (using icmppingloss) you could generate a similar graph. It won't be the same, but you could do something like it using the "bar graph" type. The one thing Zabbix should not be able to do, graph-wise, is the color coding of different levels of ping loss. Other than that, you should be able to match fairly closely.

icmppingloss[<ip>,<count>,<interval>,<size>,<timeout>]

The link you provided says it was 20 pings every 300 seconds. Try:

icmppingloss[,20,,,]
icmppingtime[,20,,,] will do the same for icmppingtime, but a single ping is possible without the bracketed part (the remaining items will all be the default values for zabbix, which you can research on your own).

Set the time between checks to be 300 seconds.

Sounds good


I'll give a try!!


thank you!!!


L.