[prev in list] [next in list] [prev in thread] [next in thread] 

List:       openvswitch-discuss
Subject:    [ovs-discuss] Packet drop on OVS-DPDK bridges
From:       vrapelly () sonusnet ! com (Rapelly, Varun)
Date:       2015-06-30 10:39:36
Message-ID: BL2PR03MB26034F17E780ECFAAEA91F3B7A90 () BL2PR03MB260 ! namprd03 ! prod ! outlook ! com
[Download RAW message or body]

Hi Team,

I created OVS bridges [on RHEL 7]using the attached Init.sh script [Using ovs-branch \
2.4 and DPDK-2.0. Not using 1GB huge pages].

Then I created a KVM instance  and used mactap devices [created by ovs bridge] in \
that VM instance. I had installed my application on new VM instance and I was able to \
do the things[calls] successfully. But when I started pumping load, observed good \
number of call failures.

Then I checked the stats on the OVS bridges and found good number drop count.

[root at artha utilities]# ./ovs-vsctl get Interface pktovs statistics
{collisions=0, rx_bytes=3020971673, rx_crc_err=0, rx_dropped=59038515, rx_errors=0, \
rx_frame_err=0, rx_over_err=0, rx_packets=13956397, tx_bytes=27949373215, \
tx_dropped=340854, tx_errors=0, tx_packets=130762243}

Please clarify:
1. What this drop count implies? If it is a problem, then how to overcome this \
problem?[ by increasing any ring or mbuf sizes??] 2. Do we need to change any buffer \
ring/size based on traffic on OVS bridges?

Regards,
Varun

-------------- next part --------------
A non-text attachment was scrubbed...
Name: Init.sh
Type: application/octet-stream
Size: 1946 bytes
Desc: Init.sh
URL: <http://openvswitch.org/pipermail/discuss/attachments/20150630/e58a183d/attachment.obj>



[prev in list] [next in list] [prev in thread] [next in thread]