home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!mcsun!uknet!str-ccsun!str-va!ccas20
- From: ccas20@vaxa.strath.ac.uk
- Newsgroups: comp.sys.sgi
- Subject: R3000 ethernet problems
- Message-ID: <1992Nov12.115136.1@vaxa.strath.ac.uk>
- Date: 12 Nov 92 11:51:36 GMT
- Sender: news@ccsun.strath.ac.uk (News account )
- Organization: Strathclyde University VAX Cluster
- Lines: 78
- Nntp-Posting-Host: vaxa
-
-
-
- INDIGO R3000 ETHERNET PROBLEMS
-
- We have always had problems with ethernet performance with the INDIGOs
- on out network, no other machines experience the same difficulties.
- The promiscuous mode fix cured 90% of the trouble but at certain times
- performance still dropped to near zero. However now that we have
- upgraded to 4.0.5 we have a new symptom. Every so often (twice a week)
- my users will complain that every thing has come to a halt. Running
- netstat reveals that one of the INDIGOs (client or NFS server) has
- stopped transmitting and receiving. There appears to be no common
- denominator as to what causes the halt, we are not running any home
- grown software that uses network communications and the installation
- is unmodified "straight from the box". It is easy to restart the
- ethernet with :
-
- ifconfig ec0 inet indigo up
-
- and instantly all is well again. (until the next time). Suprisingly
-
- ifconfig ec0 inet indigo down
-
- also restarts communications, but I wouldnt guarantee the the syntax
- is correct!
-
- If I turn network debugging on, there is a sready stream of errors
- logged on the consul. This is what I found this morning:
-
- Only local loggings work and NIS and NFS service is dead, the
- following messages are logged on the consul.
-
- >> NIS v.2 server not responding for domain "abacus"; still trying
-
- Run netstat to see what up
-
- >>alpha: 1#netstat 5
- >> input (ec0) output input (Total) output
- >> packets errs packets errs colls packets errs packets errs colls
- >>23796550 889998 95143 0 90766 23824154 889998 122747 0 90766
- >> 0 0 0 0 0 0 0 0 0 0
- >> 0 0 0 0 0 0 0 0 0 0
- >> 0 0 0 0 0 0 0 0 0 0
-
- Nothing is going in or out so toggle the ethernet.
-
- >>alpha 2# ifconfig ec0 inet alpha up
-
- And re-run netstat.
-
- >>alpha 3# !nets
- >>netstat 5
- >> input (ec0) output input (Total) output
- >> packets errs packets errs colls packets errs packets errs colls
- >>23796570 889999 95147 0 90771 23824182 889999 122759 0 90771
- >> 2827 33 3 0 16 2834 33 10 0 16
- >> 3677 35 3 0 17 3681 35 7 0 17
-
- So everything is running again and the following errors immidiately
- start to appear.
-
- >>alpha 4# ec0: CRC error
- >>ec0: dma stopped, rstat = 0x11, crbdp = 0x80155760
- >>ec0: dma stopped, rstat = 0x11, crbdp = 0x80155760
- >>ec0: dma stopped, rstat = 0x11, crbdp = 0x80155760
- >>ec0: too many oflows, changing modes
- >>ec0: packet framing error
- >>NIS server for domain "abacus" OK
-
- The vast numbers of errors/collisions are always present yet none of
- the other machines see them (4D25, 2400 and assorted SUNS). I know
- next to nothing about the mysteries of the ethernet so if any of the
- above means anything to anyone I should be most obliged if you can
- point it out.
-
- Thanks for any help
-
- Mike
-