> -----Original Message-----
> From: Testbed Support for GridPP member institutes [mailto:TB-
> [log in to unmask]] On Behalf Of Mark Mitchell
>
> Hi Ewan,
> What are the oddities?
It's a little hard to say at the moment. Roughly the sequence
of events is that (as you know) we had a lot of hassle with
our 8024F switches crashing when we first had them, which we
think we've now solved with the latest firmware. All the time
that we've been tinkering with the switches the 10Gb nics in
the disk servers have been idle and disconnected, with the
boxes on 1Gb temporary links instead. Since we think we've
sorted the switches out, yesterday I shifted our disk servers
back onto the 10Gb network where they were supposed to be.
Then, over night, seven of the machines with these cards (six
Dells and one Viglen) appeared to drop off the network (not all
at the same time, not all on the same switch) with the symptoms
that the interface would look ostensibly fine with basic ifconfig
and ethtool queries, but wouldn't actually shift any traffic.
Simply doing 'ifdown eth0 ; ifup eth0' was enough to get them back
- no forced module reloads, and no reboots or power-cycles either.
Given that relatively few sites are as far advanced as we are
with the move to 10Gb, and of those that are, I'm not sure any
are using quite the same cards (IIRC the Glasgow servers that
are on 10Gb are SuperMicro ones, which may or may not have
completely different nics) it seemed worth finding out whether
anyone's got the same. I think QMUL might have a few, bit I'm not
100% sure (and if they do, I think they'd be the CAT6 cable
version). Hence the "who's using these" question.
Ewan
|