Date: Fri, 30 Sep 2011 20:56:45 +0530 From: Naresh <gbal.naresh@gmail.com> To: Andrew Thompson <thompsa@FreeBSD.org> Cc: "freebsd-drivers@freebsd.org" <freebsd-drivers@freebsd.org> Subject: Re: Link aggregation using LACP Message-ID: <B8AC73B9-9845-40B2-88DB-555619578018@gmail.com> In-Reply-To: <CAFAOGNS8AXKdR=cYjh2Hccjaaxndb763hDXZVsh025dmdk31jw@mail.gmail.com> References: <161553F2-FE14-417C-94C6-11ACBD28E06E@gmail.com> <CAFAOGNS8AXKdR=cYjh2Hccjaaxndb763hDXZVsh025dmdk31jw@mail.gmail.com>
next in thread | previous in thread | raw e-mail | index | archive | help
Since my hosts LACP timeout is 30 seconds, switch is waiting for 3 LACP pkt= s, before it confirms link is down. So recovery is taking 90 secs. Will this timeout values also be negotiated between switch and host to=20 max value? If that is the case how to set freebsd LACP timeout to one second= ? Already I have set switch LACP timeout to one sec. -Naresh. On 29-Sep-2011, at 11:44 PM, Andrew Thompson <thompsa@FreeBSD.org> wrote: > On 30 September 2011 05:42, Naresh <gbal.naresh@gmail.com> wrote: >> Hi All, >>=20 >> I am working on a 10Gb Ethernet driver for FreeBsd 8.2. When I enable lin= k aggregation using LACP mode on two ports of our card, I expected load to b= e balanced on two ports(both are up). But interestingly traffic is flowing o= n only one interface. >=20 > Have a look at lagg_hashmbuf() which generates the hash number to > choose the port. It uses the ethernet and ip headers for this so > ensure your test traffic has enough variation. >=20 >> If I administratively bring down the interface on which traffic is flowin= g, I see lot of packet loss for about 30 secs, After that it picks up. >=20 > We don't send a gratuitous arp when the links change, maybe this is > causing your switches to not update quickly. We also use the slow lacp > timer. >=20 >=20 > Andrew
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?B8AC73B9-9845-40B2-88DB-555619578018>