CCVJ@lure.latrobe.edu.au (02/26/91)
We have three RS6000s, two model 540s and one model 530, all running at the same level (03.01.0001.0003). All three are on the same sub-network, have identically configured tcpip for nameserver, netmasks, broadcast...etc, but one of the 540s periodically loses its ability to translate names to addresses for ftp and telnet. (ping it can do and "host node" returns numbers which have had to be gotten from the nameserver!). I can fix this by flushing the routing table and restarting routed. I have run telnet with netdata toggled, and get no output (by name - it works by number). I have put the names and addresses of our most frequently accessed machines in /etc/hosts and the behaviour is strange. With a name in /etc/hosts, ftp or telnet will work - but with a terribly long delay, long enough to make most users ^C out of it. However, using the address, the response is virtually immediate. (Is this because it tries to resolve the name first?) Has anyone seen this sort of behaviour before, or can any network gurus point me in any directions? Thanks, Vicki Jordan Computing Services La Trobe University Bundoora Vic 3088 Australia ccvj@lure.latrobe.edu.au
gmoff@ccu1.aukuni.ac.nz (Moffat) (02/27/91)
CCVJ@lure.latrobe.edu.au writes: >Has anyone seen this sort of behaviour before, or can any network >gurus point me in any directions? I have seen enough of it in the last three days to want to go to the gym to develop enough strength to throw our 730 out the window *B^) >We have three RS6000s, two model 540s and one model 530, all running >at the same level (03.01.0001.0003). All three are on the same >sub-network, have identically configured tcpip for nameserver, >netmasks, broadcast...etc, but one of the 540s periodically loses its >ability to translate names to addresses for ftp and telnet. (ping it >can do and "host node" returns numbers which have had to be gotten from >the nameserver!). We have two 320s & a 730 with two ethernet interfaces, all at 3001, same mask, no nameserver. I'm trying to get the 730 to act as a router to the 320s on a subnet. The 320s appear OK but the 730 keeps exhibiting these symptoms but not always immediately - I have not yet found anything hard to cause it. >I have run telnet with netdata toggled, and get no output (by name - >it works by number). I have put the names and addresses of our most >frequently accessed machines in /etc/hosts and the behaviour is strange. >With a name in /etc/hosts, ftp or telnet will work - but with a terribly >long delay, long enough to make most users ^C out of it. However, using >the address, the response is virtually immediate. (Is this because it >tries to resolve the name first?) Using netstat -r (as smit does) to display the routing tables hangs for about 8 minutes, whereas netstat -rn is immediate. Sometimes route -f also hangs (I don't know for how long, I've always ^C'ed it) Doing a rmdev -l inet0 -d' will cure the hang (I'm not sure what it's doing, exactly, but I'm into some desperate hacking) All the hosts names are in /etc/hosts. -An aside: I have noticed that using smit to add hosts on occaisions will produce a corrupted /etc/hosts - extra or misplaced comments, from memory. My host name is the name associated with the subnet address, not the main net address, I was wondering about this. Also I have been adding/deleting routes directly with route, should I perhaps be using smit totally? (I have not yet experimented with these, I'm taking some sanity time to read the news) > I can cure the problems by flushing the routing tables and restarting >routed My problems occur whether routed is running or not! A direct question: should I specify the -g (gateway) parameter when starting routed? I have RTFM, the 'How to configure routed' man page is about as useful as tits on a bull - its a paraphrase of the routed man page. >Thanks, Me too -- Graeme Moffat, Phone : +64 9 737 999 x8384 Computer Aided Design Centre, Fax : +64 9 366 0702 School of Engineering, Mail : Private Bag, Auckland, NZ University of Auckland Email : g.moffat@aukuni.ac.nz
graeme@ccu1.aukuni.ac.nz (Graeme Moffat) (03/05/91)
Well, I've solved my problems with a little help. The hangs were due to trying to resolve names without a nameserver! (7 minutes is a ridiculously excessive timeout - where is this configured?) I had an empty resolv.conf instead of no resolv.conf - I'm sure everyone knows the difference - and no local nameserver. Earlier on I had filled in the nameserver entry of the smit tcpip minimum configuration menu, then later removed it using 'further configuration..name resolution..' when I realised I didn't want it, without knowing that the file had been created. *Flame on* This is another example of smit (I know, it's only a command interface) doing unknown things behind my back. My previous unix administration experience is zilch, and I have a lot to learn, but I can read manuals (if we had any *B^(, and I don't want things hidden away in the bowels of some festering object database, I want to know exactly what I have configured so I can take it away again if it's not right. I have poked around in a traditional BSD machine, and the config files seem fairly straightforward (not counting sendmail *B^) once I had an overview of the system (with help from a SUN installation manual - some manufacturers know how to produce readable material with examples, unlike infoexplorer, which is a dog) *Flame off* Please excuse my ramblings, but this alone has wasted two whole days of my time, and I've had others like it. -- Graeme Moffat g.moffat@aukuni.ac.nz \ Time wastes us all, Computer Aided Design Centre, Fax: +64-9-366-0702 / our bodies & our wits School of Engineering, Ph: +64-9-737-999 x8384 / But we waste time, University of Auckland, Private Bag, Auckland, NZ \ so time & we are quits
murphy@ibma0.cs.uiuc.edu (michael r murphy) (03/06/91)
My apologies to those concerned with network bandwidth, but I must say: AMEN!!! to the previous flame.