We have a machine running Scientific Linux 6.1 (RHEL recompile)
with 2.6.32-131.21.1.el6.x86_64 kernel and bridge-utils-1.2-9.el6.x86_64
It's a 12-core Xeon.
It has a Mellanox Technologies MT26448 10g NIC and
an Emulex Corporation OneConnect 10Gb NIC and bridges between them.
We see occasional packet loss through the bridge in production, which
appears to get worse when the link is busy, though it is nowhere near the
10g capacity.
If I do a flood ping through the bridge, there is no significant loss
for about 10-30 seconds, then bursts of packets are lost every second
I have adjusted kernel tuning in accordance with e.g.
http://fasterdata.es.net/fasterdata/host-tuning/linux/
and the Mellanox and Emulex user guides. It seems to make no difference to the
behaviour.
Any suggestions ?
Has anyone seen anything similar ?
--
Andrew Daviel, TRIUMF, Canada
Tel. +1 (604) 222-7376 (Pacific Time)
Network Security Manager
_______________________________________________
Bridge mailing list
Bridge@xxxxxxxxxxxxxxxxxxxxxxxxxx
https://lists.linuxfoundation.org/mailman/listinfo/bridge
