On Thu, 30 Jun 2005, [UTF-8] Jürgen Keil wrote:
Dragan Cvetkovic wrote:
I have noticed quite a few
carrier_errors on the server
bash-3.00# kstat -n elxl0 | grep carrier
carrier_errors 842217
but how came it affects NFS but doesn't affect e.g.
ftp?
Some time ago I had similar NFS problems with a tyan box with an
integrated 3com 3c920 NIC. What I found out is that full duplex does
not work very well when a certain Netgear 100mbit ethernet switch was
used. Using a Lantronix switch the problem was gone.
I know about these, I have fighting 3com cards issues for quite a while
now (and for quite a few Solaris versions). In fact, since my internal
3com card more and more often wouldn't initialise at all, I simply put
another NIC (again 3com 905 one), but without much success judging by the
number of carrier errors. Unfortunaltelly, I didn't follow carrier_errors
with my old card, but I don't remember ever having such issues where
copying a few MB file over NFS takes 10-20 minutes! I usually use scp and
that AFAICT works fine. Or better.
AFAIR, the workarounds were:
- force half duplex in /kernel/drv/elxl.conf
- use a different ethernet switch
Quite a few people have reported full duplex problems with the elxl
driver; maybe there's a bug somewhere in the elxl driver with full/half
duplex negotiation?
I have tried forcing half duplex 100Mbps in /kernel/drv/elxl.conf, but it
didn't help. With the old card I remember even booting DOS and using
some dos based 3com utility to change mode to half duplex. As for the
switches, I think we are using 3com ones.
Thanks for the suggestions.
Bye, Dragan
--
Dragan Cvetkovic,
To be or not to be is true. G. Boole No it isn't. L. E. J. Brouwer
_______________________________________________
opensolaris-discuss mailing list
opensolaris-discuss@opensolaris.org