On Mon, 16 Jan 2006, Jefferson Ogata wrote: > I certainly don't wish to offend any Sun aficionados; I myself have used > them since 3/60 days. I can certainly imagine that, to fix an issue with > an old switch, someone stuck a few lines in /etc/system; then he swaps > out the switch and those forgotten lines are no longer needed but > remain--as a result, the Sun appears not to negotiate with the new > switch because of an ancient workaround. I have no doubt that this > scenario has contributed unfairly to perceptions about Sun.
Exactly! I've had machines where the on-board hme negotiated properly, but an Sbus hme (or combo card) wouldn't - talking to the same switch! I've never had a problem with 100Mbit qfe cards - but have a 10Mbit quad card that sometimes has to be un-plumbed and re-plumbed when the cable is unplugged! Argh! And I have gotten that weekend phone call from former clients who moved a machine to a new data center and switch who forgot to look in /etc/system and got burned by auto-negotiation problems... ah, well. We call that "job security..." > I note, however, that even a new HP Procurve 4108gl switch I purchased > within the last three years required hardwiring duplex when I connected > it to Sun E5500 and E250R boxes at 100Mb/s, while having no trouble with > other boxes (mostly Intel). So I don't think all the problems have been > completely resolved. What's funny, though, is that nearly all those problems I described above were with a Cisco Catalyst 5000 - but the same interfaces plugged into a 3548 worked just fine! Over on the NetApp toasters list folks are still running into problems with duplex issues - things "appear" fine at first blush, but NFS performance sucks; lo and behold, it's a bad switch or a cranky interface or someone tweaked the startup files to hardwire a port... So, I agree with Caspar, certainly no one vendor is to blame. -- Chris
