On February 25, 2008 01:46:04 pm Richard Daemon wrote: > On Sun, Feb 24, 2008 at 9:33 AM, Vijay Sankar <[EMAIL PROTECTED]> wrote: > > Good day, > > > > I have two interfaces -- nfe0 on switch0 and nfe1 on switch1 are part of > > trunk0. Trunk failover from nfe0 to nfe1 works very well. No problems > > if switch 0 goes offline -- traffic goes through switch1 flawlessly. Once > > switch0 comes back online, traffic is disrupted for about 30 seconds. > > > > I would like traffic to continue through switch1 after switch0 is back > > online (or at least have a delay of 30 or 45 seconds before failing back > > to the master) and don't know how to do this. Is this possible? Should I > > be using ifstated for this in addition to trunk? > > > > Please let me know of any clues to resolving this. > > > > Thanks very much, > > > > Vijay > > > > > Hi Vijay, > > I don't know if anyone responded to you on this but I am very curious > to know myself as well... > Personally, I haven't used trunk(4) too much yet so I might not be of much > help. > > My guess would be either something with the rules (with regards to > keeping state or finding a way to sync the states) unless that 30 > seconds is normal??? > But to me that seems odd to have that long of a disruption, 30 seconds, > ouch. > > The other option you said was to delay it 30-45 seconds. > For that, then I would personally think that ifstated would or could > do the trick, but maybe someone else can give better feedback than me > on this whole issue? > > Also, could it be caused by something with the switches layer 2 cache > timeout period or something to that effect? Just a thought. > > Regards, > > Richard
Thanks very much for your reply. The Cisco switches have STP enabled but not RSTP. Basically it looks like when a switch comes back on line, it takes close to 30s before the port is active (meaning orange light turning to green for the port) but as far as the NIC is concerned, as soon as it detects that the link is up, the master seems to want to take over from the active port. The problem I have is that people are comparing the NIC "teaming" on Windows Servers to OpenBSD's "trunking". With teaming, there is no preference for either members of the "team" meaning when a switch comes back, they don't notice the disruption since the port doesn't fail back to the master. The IT guy has to manually do that if necessary. But with trunking since the master is always preferred for traffic, the 30s delay as a result of whatever is going on in the switch is noticed. Anyways, hopefully someone more knowledgeable than us will notice this thread and suggest something that I can try. At this moment, I don't know what is the right place to put my effort in. Thanks again, Vijay -- Vijay Sankar, M.Eng., P.Eng. ForeTell Technologies Limited 59 Flamingo Avenue, Winnipeg, MB Canada R3J 0X6 Phone: +1 204 885 9535, E-Mail: [EMAIL PROTECTED]
