Regards _Sugesh
From: qintao (F) [mailto:qint...@huawei.com] Sent: Friday, July 22, 2016 8:09 AM To: Chandran, Sugesh <sugesh.chand...@intel.com> Cc: wuhao 00271188 <w00271...@notesmail.huawei.com.cn>; guoyilong 00231009 <g00231...@notesmail.huawei.com.cn>; lukai 00197897 <l00197...@notesmail.huawei.com.cn>; liuguifeng 00230956 <l00230...@notesmail.huawei.com.cn>; qintao (F) <qint...@huawei.com>; wangmeiling 00346626 <w00346...@notesmail.huawei.com.cn>; kangxu 00360128 <k00360...@notesmail.huawei.com.cn> Subject: a question about userspace ovs Dear sugesh, Recently I have met a problem when I do some tests. The first following picture is my network structure. As the second and third following picture show , the problem is that the vm(192.168.0.16) cannot ping another vm(192.168.0.19) successfully when being adopted the big size packets ,such as the mtu 1500 . But after being adopted the small size packets ,such as the mtu 64 ,the vm(192.168.0.16) can ping another vm (192.168.0.19)successfully. And then I chose the netperf as my test tool .Whatever the size of the packets I have changed , the packets generated by the netperf have been dropped fully by the port dpdk ,when I run the command "watch -d -n 1 ovs-ofctl dump-ports br0".And the I found the numbers of free hugepages on my host is zero. So I have a doubt it if the problem has something to do with it ,or other reasons? [Sugesh] Looks to me that the MTU on your interfaces are not sufficient to handle the 1500 traffic because every vxlan traffic adds up 50 bytes of additional overhead. Best regards, Tony QIN [cid:image001.png@01D1E43A.0D796270] [cid:image003.png@01D1E43A.0D796270] [cid:image004.png@01D1E43A.0D796270] [cid:image005.png@01D1E43A.0D796270]
_______________________________________________ discuss mailing list discuss@openvswitch.org http://openvswitch.org/mailman/listinfo/discuss
