HANK@BARILVM.BITNET (Hank Nussbacher) (08/19/90)
I have defined two ciscos (8.1(14)) each with bridging between the the two of them as well as routing. I have two parallel 19.2kb lines between the two of them defined in a circuit group so as to perform load balancing. I open a LAT connection from a Decserver 200 over to a Vaxstation 3100 and start typing a file. The output is all being bridged. When I remove one of the lines physically from the router, the connection dies between the Decserver and the Vaxstation. This does not happen when I do a connect from one Vax station to another (using routing rather than bridging) and removing one of the cables (physically). If I do not have the 2 lines defined as a circuit group then the connection does immediately. If it is defined within a circuit group, the connection can either disconnect immediately if the line I disconnect is the line that is being transmitted on or it can disconnect in 10-15 seconds if the line I am transmitting on (serial 0) is still connected but the bridge at the other side is transmitting on serial 1 which is now disconnected. It may just be that I am don't understand circuit load balancing enough but I think this is a bug. Anyone seen this? Thanks, Hank
lougheed@cisco.com (08/20/90)
Hank - From your message I am unable to ascertain if you have a cisco problem or not. LAT is notoriously delay sensitive; its timers may be expiring before the cisco spanning tree timers are kicking in. After your connection breaks, are you able to re-establish a connection? If so, this would indicate either a super-sensitive LAT, a slow cisco, or something else. It would be helpful if you experimented a bit more along these lines and communicated your results to the "customer-service@cisco.com" mailing list. It would help if you included configurations and information on the bandwidth of the serial lines. As for the IEEE spanning tree protocol, yes, 8.1(14) has some problems in that regard. There is the serial line crash you discovered, as well incorrect multicast addresses. These problems have been resolved in the current field release, 8.1(19). Kirk Lougheed cisco
SHERWOOD@AC.DAL.CA (John Sherwood) (08/21/90)
Hank: >I have defined two ciscos (8.1(14)) each with bridging between the >the two of them as well as routing. I have two parallel 19.2kb lines >between the two of them defined in a circuit group so as to perform >load balancing. I open a LAT connection from a Decserver 200 over to >a Vaxstation 3100 and start typing a file. The output is all being bridged. >When I remove one of the lines physically from the router, the connection >dies between the Decserver and the Vaxstation. LAT is a very time-sensitive protocol. I would not be surprised to find it will not work on a single 19.2k link. I am surprised that you can get it to work at all even with 2 links sharing the load. Even minor delays in packet delivery can give LAT heartburn. > This does not happen >when I do a connect from one Vax station to another (using routing rather >than bridging) and removing one of the cables (physically). I am guessing that you are using DECnet CTERM rather than LAT. CTERM (ie. $SET HOST xxxx) should work with no problems over slow links; we run it at 4800 regularly. I would try switching to CTERM as the terminal protocol and see if your problems go away. Cheers John Sherwood Dalhousie University Halifax, Nova Scotia sherwood@ac.dal.ca.
dss@unix.cis.pitt.edu (Don Salvin) (08/22/90)
Newsgroups: comp.dcom.sys.cisco Subject: Re: Bug in cisco bridging? Summary: References: <24936@boulder.Colorado.EDU> Sender: Reply-To: dss@unix.cis.pitt.edu (Don Salvin) Followup-To: Distribution: Organization: Univ. of Pittsburgh, Comp & Info Services Keywords: In article <24936@boulder.Colorado.EDU> SHERWOOD@AC.DAL.CA (John Sherwood) writes: >Hank: > >>I have defined two ciscos (8.1(14)) each with bridging between the [stuff deleted...] >>dies between the Decserver and the Vaxstation. > >LAT is a very time-sensitive protocol. I would not be surprised to find >it will not work on a single 19.2k link. I am surprised that you can get it >to work at all even with 2 links sharing the load. Even minor delays >in packet delivery can give LAT heartburn. It is true that the DEFAULT timers provided with the DECserver 200's make them quite sensitive to "minor" delays in the network. However, they are parameters in the server that are settable. At our site we run MANy servers, some of them running over slow, crowded links, and have no problems. This is admittedly a pain to administer in a large network, but we have written tools to make it easier for us. The timers you want to set are RETRANSMIT LIMIT and KEEPALIVE TIMER. We use RETRANSMIT LIMIT=60 and KEEPALIVE TIMER=180 . >> This does not happen >>when I do a connect from one Vax station to another (using routing rather >>than bridging) and removing one of the cables (physically).> >I am guessing that you are using DECnet CTERM rather than LAT. CTERM >(ie. $SET HOST xxxx) should work with no problems over slow links; we run >it at 4800 regularly. > >I would try switching to CTERM as the terminal protocol and see if your >problems go away. I would try fixing the timers first, as LAT is more efficient in its use of network bandwidth, something I would be concerned about, especially over slow links. It also provides better functionality as a terminal protocol than CTERM. >Cheers > >John Sherwood >Dalhousie University >Halifax, Nova Scotia > >sherwood@ac.dal.ca. Don Salvin Data Comm Manager, University of Pittsburgh dss@pitt.edu