Z
Z
zorghrd2016-08-07 22:35:51
linux
zorghrd, 2016-08-07 22:35:51

How to determine the source of a network problem in CentOS?

There is physics, there are several LXC containers on it.
Physics and almost all containers have 1 IP address.
One of the containers has a separate IP address with a separate MAC.
Recently, out of the blue, it crashed.
The main IP address stops responding, is unavailable on any ports, respectively, at this moment all containers are unavailable on the same IP.
At the same time, the container that sits on a dedicated IP remains fully operational, its IP is working, the container is available.
Usually, before dumping the main IP, an error occurs several times:
kernel: unregister_netdevice: waiting for lo to become free. Usage count = 1
I sinned against the container, from which a lot has been pouring into dmesg lately:
unexpectedly shrunk window dmesg
But disabling this container didn't help. The network on the main IP still falls off.
Immediately after rebooting the hardware, the network works correctly for several minutes, but then it starts to disappear, then reappear.
Server on CentOS 7.
Where to dig?

Answer the question

In order to leave comments, you need to log in

1 answer(s)
Z
zorgrhrd, 2016-08-11
@zogrhrd

In general, I removed the network from the bridge, demolished it completely, created a new one, added it to the same bridge, worked like clockwork for several days, now it fell off again for 3 minutes. At the same time, as before, both other interfaces on the same bridge worked without failures, without packet loss. I checked the monitoring, no anomalies, all the I / O graphs are smooth, on the CPU, on the RAM, on the network and disks, the only 100% packet loss on this network interface at the time of the fall.
I don't know where to dig anymore, I'm going to move to another server.

Didn't find what you were looking for?

Ask your question

Ask a Question

731 491 924 answers to any question