Matt Harrison wrote: > Robert Milkowski wrote: >> Hello Matt, >> >> Tuesday, September 30, 2008, 10:52:57 PM, you wrote: >> >> MH> Everything was working fine on our snv_97 fileserver then one day our >> MH> transfers had slowed to a crawl. I originally thought it was CIFS only >> MH> but after some tests I can see that its all network transfers including >> MH> scp and ftp. >> >> MH> I've run some benchmarks like bonnie++ and disk IO seems to be ok. >> >> MH> A zpool scrub doesn't bring any errors and takes around the same time as >> MH> normal. >> >> MH> The problem is that copying to the machine will run at about 20Kb/s and >> MH> from the machine is faster, somewhere in the region of 60-100Kb/s. >> >> MH> What on earth could be causing this massive slowdown? I've rebooted the >> MH> machine, checked all the processes, checked for faulty devices, >> MH> everything I can think of. >> >> MH> Any ideas? >> >> Network issues? >> >> > > Actually this could be related. About a week ago, the onboard NICs > started playing up and would drop connection about 70% of the time. The > only solution after talking with the list was to put in a PCI NIC, and > it's possible that it might have some issues (it's really quite old). > Could packet fragmentation on the NIC be causing this but not bringing > down the whole connection?
Just replaced the old NIC with a nice shiney new RTL8139 and it works beautifully, speed to and from is now back to what it should be. I don't know why I didn't associate the slowdown with the NIC change in the first place, it should've been the first thing I looked at. Thanks for the suggestions. Matt _______________________________________________ opensolaris-discuss mailing list opensolaris-discuss@opensolaris.org
