I was sending 5000 syslog messages per second to a virtual machine running on a
CentOS 5.2 host, I have a program running on the virtual machine that receives
and processes the syslog messages. What I observed is that there is about 30%
packets dropped by the dom0 peth0 interface. During the test period, the CPU
usage on the virtual machine is about 30-40%. I already did : echo 256 >
/sys/class/net/eth0/rxbuf_min on the domU.
Does anyone know how this happen?
Thanks,
Hongbo
_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users
Quoting Hongbo Zhu <hongbo_zhu@yahoo.com>:> I was sending 5000 syslog messages per second to a virtual machine > running on a CentOS 5.2 host, I have a program running on the > virtual machine that receives and processes the syslog messages. > What I observed is that there is about 30% packets dropped by the > dom0 peth0 interface. During the test period, the CPU usage on the > virtual machine is about 30-40%. I already did : echo 256 > > /sys/class/net/eth0/rxbuf_min on the domU. > > Does anyone know how this happen? > > Thanks, > > HongboIn ''dmesg'' are you seeing connection tracking full dropping packets or something similar? Jon _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
dmesg on both domU and dom0 showed nothing related to packet dropping. If I do "ifconfig peth0" on dom0, I can see the number of dropped packets keep increasing. ________________________________ From: "jonr@destar.net" <jonr@destar.net> To: xen-users@lists.xensource.com Sent: Friday, October 24, 2008 5:34:19 PM Subject: Re: [Xen-users] udp packet loss Quoting Hongbo Zhu <hongbo_zhu@yahoo.com>:> I was sending 5000 syslog messages per second to a virtual machine > running on a CentOS 5.2 host, I have a program running on the > virtual machine that receives and processes the syslog messages. > What I observed is that there is about 30% packets dropped by the > dom0 peth0 interface. During the test period, the CPU usage on the > virtual machine is about 30-40%. I already did : echo 256 > > /sys/class/net/eth0/rxbuf_min on the domU. > > Does anyone know how this happen? > > Thanks, > > HongboIn ''dmesg'' are you seeing connection tracking full dropping packets or something similar? Jon _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
Quoting Hongbo Zhu <hongbo_zhu@yahoo.com>:> dmesg on both domU and dom0 showed nothing related to packet > dropping. If I do "ifconfig peth0" on dom0, I can see the number of > dropped packets keep increasing. >Is this the only domu on the box? Maybe dedicate a nic to that domu. What are the values of the below commands on both boxes? cat /proc/sys/net/ipv4/netfilter/ip_conntrack_max grep -c ^udp /proc/net/ip_conntrack grep -c ^tcp /proc/net/ip_conntrack Jon _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users