Can't monitor LAN when Internet connection lost

[zabbixchreese]

Hello,

We are using Zabbix live CD/applicnace 1.8.11 on a laptop plugged into the LAN (behind our router) or our network. I've set it up to help track down some network Internet problems (trying to determine if problem is within the LAN or outside) using the icmp ping tests (pinging places outside on the internet, other computers on the LAN, router LAN gateway, WAN gateway, etc.).

What I have noticed is that if I pull the network cable out from the WAN of our router to simulate an Internet outage, the zabbix graphs show everything down including things on the LAN side. "down" being defined as me seeing gaps in the ping graphs. When I look at the raw data values, there are missing data points (if we ping every 30 seconds) sort of like the following:
12:01:00 3.000
12:05:00 2.300
12:05:30: 3.100
12:06:00: 3.000
12:06:30: 3.200

etc.

I am pinging based on IP address and not DNS. On an older version of zabbix that we run elsewhere, I usually see 0.00 values. Here it is just skipped.

Not sure what is going on. Could there be some kind of load problem where zabbix is freaking out when it can't reach the servers it is trying to ping out on the Internet, or is there some kind of DNS issue so that if zabbix itself cannot reach the Internet, it makes everything down?

Thanks for any help.

[nelsonab]

Sounds like you've got a routing issue going on somewhere. Perhaps your router is relying on something on the WAN side and thus blocks when the WAN goes down? If that is the case that would make sense for you to see intermittent drops due to jitter.

You're probably going to need to go through your router's configuration bit by bit, I suggest writing it down on paper by hand as you do this. Do not use copy/paste, you need to pay attention to all details, copy/paste allows you to gloss over things.

You will also probably need to get a packet sniffer on your network. Tcpdump has some options which can allow you to capture to a file or a specified length of time or size and to keep the previous N copies before deleting.

In addition if you're running this in a VM on a laptop that may very well be part your problem as well. Have a look at the load of the VM, what is it's wait state? Laptops typically have SLOW hard drives and are almost NEVER suited for relational databases, which Zabbix requires. If you *MUST* run it on the laptop, install Linux on the bare metal, and run Zabbix as close to the hardware as you can. If you can't do that, consider giving the VM more resources. In addition check the load on the hypervisor, if you're seeing that the hypervisor is busy it's possible your VM is being starved of CPU cycles, which in some cases can leave to a kernel panic.

Unfortunately your description is light in background and heavy on numeric numbers which have no meaning or use without context. Without knowing more my guess is your issue will be a combination of two or more seemingly separate but related issues.

[zabbixchreese]

Thanks for the suggestions. I can ping using the command line/console on the zabbix machine to the other LAN desktop without problems when I create the WAN outage. Not sure if a major difference between fping and ping. I was using ping.

I believe I had the same problem with another laptop I'm running with the same setup placed outside the router sitting on the same network wan segment as the router. If I disconnect the network cable going to the ISP router, my pings to the router WAN shows gaps.

I'm not running it on a VM. It's right off of the live CD. There are some built-in auto-installed tools that monitor the zabbix server itself that don't seem to report any problems w/ load, etc.

I was hoping to use zabbix as sort of a portable network diagnostic tool where I could temporarily strategically setup (inside LAN, outside WAN, etc.) as needed whenever a client had some problems diagnosing network problems.