On Tue, 2 Feb 2021 at 11:19, BERTRAND Joël <joel.bertr...@systella.fr> wrote: > > RVP a écrit : > > On Tue, 2 Feb 2021, BERTRAND Joël wrote: > > > >>> 1. Are ethernet cards on either end gigabit adapters? > >> > >> Yes. > >> > >>> 2. Are the cards configured as such by their respective OSes? > >>> (use ifconfig -v ifname) > >> > >> Yes. > >> > >>> 3. Is the ethernet cable between the two machines Cat-5E or > >>> better (with all 8 wires)? (Performance is sometimes not up > >>> to spec with some cards and mere Cat-5 cables.) > >> Cat6a... > >> > > > > Any errors on the interfaces? > > Use: netstat -I ifname > > on NetBSD. > > No, I have verified before my first message : > > legendre:[~] > netstat -I wm0 > Name Mtu Network Address Ipkts Ierrs Opkts > Oerrs Colls > wm0 9000 <Link> b4:96:91:92:77:6e 21863327 0 90733314 > 0 0 > wm0 9000 192.168.12/24 legendre.systella 21863327 0 90733314 > 0 0 > wm0 9000 fe80::/64 fe80::b696:91ff:f 21863327 0 90733314 > 0 0 > legendre:[~] >
Looking at one of my systems, I see I have identical setups for iSCSI served by in-kernel iscsi target and for net/istgt, backed by zvols. A few years ago I started with the built-in one, but had a number of problems accessing from Windows 10 and Windows Server machines. I moved the setup to net/istgt and haven't had any problems since. I don't remember all the reasons I did that, but they included the clients not attaching after a reboot or hibernation and targets disappearing after a reboot of the NetBSD server; I've had no such problems with the package. I suppose the user-level package probably is less efficient etc., but it has worked fine for me and perhaps it's worth trying. Chavdar -- ----
