On 6/14/2011 7:42 PM, [email protected] wrote: > On Tue, Jun 14, 2011 at 10:12:45PM -0700, Dave Close wrote: >> This box has two on-board Broadcom NICs >> and we added an Intel quad-port NIC sold by Dell. > ... > >> Strange interactions should >> be expected if the combination hasn't been explicitly tested. > > Yeah, that's why people do things like pay extra to get a dell-branded > intel nic. Personally, my experience, like the one above, has been that > it doesn't seem to help much. paid support is great for obvious (box > won't boot, etc...) problems but is generally worthless for subtle stuff. > > It's irritating, 'cause I imagine dell could do a much better job of > integration testing than I can. > > I've got a pair of PowerEdge SC1425 in my environment, identical boxes set-up as an HA pair, with an Intel 82541GI adapter in them. Both are hit by the lovely "e1000: eth0: e1000_clean_tx_irq: Detected Tx Unit Hang" bug. Every few months I try the latest network driver module from Intel in the vain hope it'll finally fix the problem, but always to no avail. Can't predict when it will happen, so I've never managed to catch the event in a tcpdump, and it always happens for barely more than a couple of seconds, but it's enough to trigger a failover.
It's far from the first quirk I've seen with hardware by various vendors running supported and 'tested' OSs. I've long since reached a stage where I just don't expect any hardware to work glitch free. In some regards that's probably not a bad thing, it encourages thinking about things decentralised. Service uptime, not server uptime, after all :) Paul _______________________________________________ Tech mailing list [email protected] https://lists.lopsa.org/cgi-bin/mailman/listinfo/tech This list provided by the League of Professional System Administrators http://lopsa.org/
