Howdy, Stranger!

It looks like you're new here. If you want to get involved, click one of these buttons!


Packet drops on KVM guest
New on LowEndTalk? Please Register and read our Community Rules.

All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.

Packet drops on KVM guest

BochiBochi Member

Hey guys,

got aware of this rather strange behaviour of a KVM guest this day:

In plain words:
There seems to be a pretty constant drop rate on my bridged network interface of the KVM guest,
but on the host node everything seems fine.
Tried to ping for a few minutes and used the service provided by this box -> no problems.
I found some bug reports online regarding a similiar problem, but that dated back a few years and where related to an older kernel version...

Maybe this is a rather common problem and I am just to inexperienced - do you have any suggestions?

Regards, Bochi

Comments

    1. Check your host's irq balancing - maybe all the interrupts happen on a single overloaded core
    2. Check libvirt's config for traffic shaping
    3. Make sure you don't have any tc rules set up
    4. Check your guest's interface load / pps to make sure there is no flooding
    Thanked by 1Bochi
  • BochiBochi Member
    edited July 2015

    @yourserverse said:
    1. Check your host's irq balancing - maybe all the interrupts happen on a single overloaded core

    root@node01:~# grep eth /proc/interrupts
     48:    5016262          0          0          0   PCI-MSI-edge      eth0

    All interrupts are handled by one core, but that should be fine - affinity is set accordingly.
    And load on the systems (node and guest) is well below 0.1, so this shouldn't be a problem.

    1. Check libvirt's config for traffic shaping

    There should be no QoS in place, checked the guest's XML.

    1. Make sure you don't have any tc rules set up
    root@guest:~$ tc qdisc show
    qdisc pfifo_fast 0: dev eth0 root refcnt 2 bands 3 priomap  1 2 2 2 1 2 0 0 1 1                   1 1 1 1 1 1

    Seems alright to me, same on host.

    1. Check your guest's interface load / pps to make sure there is no flooding
    root@guest:~/scripts$ ./show_pps.sh eth0
    TX eth0: 6 pkts/s RX eth0: 10 pkts/s
    TX eth0: 10 pkts/s RX eth0: 10 pkts/s
    TX eth0: 8 pkts/s RX eth0: 11 pkts/s
    TX eth0: 8 pkts/s RX eth0: 10 pkts/s
    TX eth0: 9 pkts/s RX eth0: 13 pkts/s
    TX eth0: 7 pkts/s RX eth0: 10 pkts/s
    TX eth0: 2 pkts/s RX eth0: 3 pkts/s

    Guest is mostly idling, but checked that too: no flooding...

    Thank you very much for your effort so far! :)

    Thanked by 1yourserverse
  • BochiBochi Member

    Checked the actual ratio and it seems to be quite low:

    vmbr0     Link encap:Ethernet  HWaddr xx:xx:xx:xx:xx:xx
              inet addr:xxx.xxx.xxx.xxx  Bcast:xxx.xxx.xxx.xxx  Mask:255.255.255.0
              UP BROADCAST RUNNING MULTICAST  MTU:1500  Metric:1
              RX packets:5007634 errors:0 dropped:484 overruns:0 frame:0
              TX packets:501751 errors:0 dropped:0 overruns:0 carrier:0
              collisions:0 txqueuelen:0
              RX bytes:860548456 (860.5 MB)  TX bytes:65380152 (65.3 MB)

    But nevertheless it somehow annoys me. :P
    Especially as long as I don't know the reason for this... Any other suggestions?

Sign In or Register to comment.