[Qemu-discuss] Sporadic input / output error on x86-64 linux guest

[Davide Baldini]

This is a forward of bug #1502095.

Bug Description:


** PROBLEM **

Sporadic, unexplained input output error.
When I try to SSH to one of the instances, most of the times everything just 
works fine. Some other times, the SSH connection can be established, but as I 
interact with the guest via SSH (ls, cd, top) the guest reports "input / output 
error" on the SSH console. Sometimes, the SSH daemon on the guest answers 
"/bin/bash input output error". Sometimes the SSH client answers that the 
connection has been dropped by the host.

When this happens, the web services I run on the VMs are and remain unreachable, 
the VMs themselves are and remain unreachable/unusable via SSH as described, and 
after killing the relative qemu processes and restarting the VMs, no recent logs 
are registered on /var/logs on the guests, arguably since the time of the crash.
The crash is not triggered by SSH, I just find SSH to be nonfunctional after I 
spot my web services offline.

This only happens to one VM at a time, independently. The other VMs appear to 
run fine when one VM encounters the problem.

I wish instructions to further debug this.


** SETUP: **

→ Host
    Qemu version 2.4.0.1
    Linux: 4.1.1 (Debian 8.2, GCC 4.9.2, x86_64)
    filesystem ext3 and ext4
    one single host machine

→ Guests (3 VMs)
    architecture x86_64, Linux 3.16.0-4-amd64 (Debian 7.6)
    virtual disk qcow2, uncompressed
    guests filesystem ext3
    virtual disks size: VM1: 3GB, VM2: 5GB, VM3: 250GB

→ Network
    bridge (br0) and tap interfaces.


** COMMAND LINE **

For all 3 VMs, the command line is similar to the following. Only RAM size and 
ID ("109") are changed.

    /usr/local/bin/qemu-system-x86_64 -hda /media/raid1/qemu-109 -m 8G -smp 4 
-enable-kvm \
        -netdev bridge,id=br109 -device 
virtio-net-pci,netdev=br109,id=nic0,mac=00:00:00:00:01:09 \
        -k it -daemonize


-- Davide