jstewart@sce.carleton.ca (John Stewart) (11/16/89)
We backup a number of Sun machines on our ethernet backbone using rdump. While this is running the network is busy but still usable by Sun machines. There are also 3 DEC machines on the backbone that are VaxCluster'd with a megaton of DEC hardware in our Physics subnet (isolated from the backbone via a Retix bridge). The DEC machines seem to intensely dislike the network load because they start crashing and rebooting themselves while we are running our backups. Our physics people blame the problem on our backbone ethernet (and us by insinuation as we maintain the backbone). I think the problem lies with the VaxCluster software but we're not a DEC shop so we can only speculate about that. If anyone else has run into this problem I sure would like to hear your advice. -- "Support the President's War On Long Usenet Signatures"
Gene.Hastings@BOOLE.ECE.CMU.EDU (11/20/89)
There are a number of simple (I hope!) ideas to pursue: (Most of them require some sort of Ethernet monitor, like PC Netwatch, FTP Software Lanwatch, TCPdump on a Sun, or a special purpose monitor like an Excelan Lanlyser, or a Network General Sniffer.) Is your Retix bridge really filtering? If it is, the Physics people should not see any non-broadcast traffic. Check the Physics cable for existence of the rdump traffic or destination addresses that aren't on it. There are two things that can be affecting the VMS hosts - inordinate amounts of broadcast traffic (which would affect any host), or VERY high traffic on the cable, this last because of features of the way DECNET routing updates are done. Chances are, iff the Vax subnet is seeing your traffic (and it can't be avoided), some help can be gained by lengthening (in NCP) the DECNET routing period. Have them check the DECNET manuals for ranges of settings for the announce and listen parameters. Gene