The Intel S3210SHLX mainboard has a pair of PCI-X slots on it, and supports
the Core 2s nicely, so that should overcome Will's limitations.
To my knowledge you cannot prevent it from binding, but you can use Target
Portal Group Tags (TPGT) to limit what IPs it will accept traffic on (on a
per-iqn basis). IIRC this does not work well with enabling iscsi via zfs.
I've found that I needed to create the backing store then iscsitadm to
create the targets and assign the portal group tags, otherwise it would
"forget" the portal group assignments on reboot. AFAIK this has not
changed, and is (IMHO) one of the more glaring remaining deficiencies in
ZFS.
If that's the worst I can complain about, it's pretty good stuff ;)
Andrew Hettinger
http://Prominic.NET || [EMAIL PROTECTED]
Tel: 866.339.3169 (toll free) -or- +1.217.356.2888 x.110 (int'l)
Fax: 866.372.3356 (toll free) -or- +1.217.356.3356 (int'l)
Mobile direct: 1.217.621.2540
CompTIA A+, CompTIA Network+, MCP
[EMAIL PROTECTED] wrote on 09/08/2008 07:35:37 AM:
> 2008/9/8 Will Murnane <[EMAIL PROTECTED]>:
> > replacing the
> > motherboard with a cheap Core 2 one significantly *improved*
> > performance, despite limiting the controller to vanilla-PCI speeds.
> > CPU usage wasn't really an issue, but speeds were nevertheless limited
> > to 20 MB/s or so locally. With the new motherboard I get about 70
> > MB/s over network, mostly disk-limited.
>
> That's interesting. Getting a new board and cheap modern cpu isn't
> really an issue so i'll most likely do that. My old fileserver was
> managing 500MB/sec burst and 300MB/sec sustained read (according to
> hdtach) so i'd not like to be going completely backwards with my new
> setup.
>
>
> > You'll have to present part of the volume via iSCSI and part via NFS;
> > they can't overlap. Also, you must define how large an iSCSI volume
> > you want up front. Although you can resize it later, this will entail
> > resizing the filesystem over network to allow the remote machine to
> > use all the new space.
>
> Sorry, I wasn't clear. I didn't mean to use both protocols on the same
> slice of storage
>
>
> > If you assign an IP address to a single NIC, then all traffic will
> > flow over that interface. Trunking Intel interfaces is also possible,
> > but if you don't want that to happen it won't by default.
>
> Yes, I understand that. However, if I have two NIC's on two seperate
> subnets and simply enable iSCSI, how do I bind it to only one of the
> NIC's? I don't want the storage to be presented at all on the other
> NIC(s). I'll go do some reading on it incase it's stupidly simple :)
>
> Thanks all.
> _______________________________________________
> storage-discuss mailing list
> [email protected]
> http://mail.opensolaris.org/mailman/listinfo/storage-discuss
_______________________________________________
storage-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/storage-discuss
