On 24 jun 2008, at 00.39, Michael Schuster wrote: > Marcus Agehall wrote: >> I'm having problems with my fileserver, which is still running the >> svn_66 build. >> When I push a lot of data from my clients to it, it seems like the >> NIC just gives up after a while and gives me the following in the >> /var/adm/messages: >> Jun 23 19:21:54 ironforge rge: [ID 801725 kern.info] NOTICE: rge0: >> link down (initialized) >> Jun 23 19:21:54 ironforge in.routed[582]: [ID 238047 >> daemon.warning] interface rge0 to 192.168.0.100 turned off >> Jun 23 19:21:58 ironforge rge: [ID 801725 kern.info] NOTICE: rge0: >> link up 1000Mbps Full_Duplex (initialized) >> Jun 23 19:21:58 ironforge in.routed[582]: [ID 300549 >> daemon.warning] interface rge0 to 192.168.0.100 restored >> Jun 23 19:22:08 ironforge rge: [ID 801725 kern.info] NOTICE: rge0: >> link up 1000Mbps Full_Duplex (initialized) >> Jun 23 19:22:08 ironforge rge: [ID 801725 kern.info] NOTICE: rge0: >> link down (initialized) >> Jun 23 19:22:08 ironforge in.routed[582]: [ID 238047 >> daemon.warning] interface rge0 to 192.168.0.100 turned off >> Jun 23 19:22:12 ironforge rge: [ID 801725 kern.info] NOTICE: rge0: >> link up 1000Mbps Full_Duplex (initialized) >> Jun 23 19:22:12 ironforge in.routed[582]: [ID 300549 >> daemon.warning] interface rge0 to 192.168.0.100 restored >> Jun 23 19:34:00 ironforge rge: [ID 801725 kern.info] NOTICE: rge0: >> link up 1000Mbps Full_Duplex (initialized) >> It seems my windows clients survives a few hickups, but after a >> little while, they choke on them and abort the transfer. The same >> problem seems >> to appear when running iperf over a period of 60s or so. >> I've been trying to track this problem down for a while now, >> basically getting nowhere. Since I can replicate the problem with >> iperf, I think I >> can pretty much conclude that it's not related to samba/zfs, but >> rather >> the rge driver or some other hardware in my system. >> Anyone have any ideas? If there is any additional information I can >> post to help track down my problems, feel free to ask. I simply don't >> know what to post, since everything I've looked at looks ok to me... > > have you looked at the switch and the cables involved? One thing you > could try is using different ports on the switch and/or different > cables (but only change one thing at a time, test, record results, > repeat ;-).
Yes. I've tried with several different 100mbit switches and Cat5/Cat6 cables. Last weekend I went out and got myself a nice new NetGear GS108 gigabit switch and a GA311 gigabit ethernet card for the server and I still get the same results. On the client side, I have three different computers running Linux, WinXP, WinXP64 and they are all able to make my fileserver disconnect if I push data thru for a while. Just for kicks, I tried to throttle the traffic from a gigabit client thru a 100mbit switch and it produced the same result, so it doesn't seem to be speed related either. It feels like some counter is wrapping after a while and causes the network driver to reset, but that is just a gut feeling and I could be terribly wrong here... Best regards, Marcus _______________________________________________ networking-discuss mailing list [email protected]
