mheffron@orion.oac.uci.edu (Matt Heffron) (09/15/90)
I'm having trouble with running /etc/named under SR 10.2. It seems to run fine for a while, but if it is idle for very long (5-10 minutes, maybe longer,...) it then fails to respond within a reasonable time to a request, and the requestor times out (~30 sec) if the request is repeated immediately, then it works ok for awhile again. Does anybody have any ideas about this? Pertinant details: DN3500 8MB w/disk Ethernet & DomainRing the node is running routed, but there are no nodes on the ring side running TCP/IP yet. The request for name resolution is from an IBM PS/2 on the same Ethernet. The routed patch m0108 is NOT yet installed (does it matter??). Thanks, Matt Heffron mheffron@orion.oac.uci.edu
jal@acc (John Lauro) (09/17/90)
In article <26F169EE.21884@orion.oac.uci.edu> mheffron@orion.oac.uci.edu (Matt Heffron) writes: > >I'm having trouble with running /etc/named under SR 10.2. >It seems to run fine for a while, but if it is idle for very long >(5-10 minutes, maybe longer,...) it then fails to respond within >a reasonable time to a request, and the requestor times out (~30 sec) >if the request is repeated immediately, then it works ok for awhile >again. Does anybody have any ideas about this? >Pertinant details: >DN3500 8MB w/disk >Ethernet & DomainRing >the node is running routed, but there are no nodes on the ring side >running TCP/IP yet. >The request for name resolution is from an IBM PS/2 on the same Ethernet. >The routed patch m0108 is NOT yet installed (does it matter??). > >Thanks, >Matt Heffron mheffron@orion.oac.uci.edu I think at least part of the problem is with the software you are running on the IBM. If it's something local that NAMED is an authority for, I always get a quick response. If it's a distant lookup, the following happens: (Using domain trace with KA9Q NOS on a PC) received message length -1 (errno 17) Nos than retries and it ussually works the second or third time. Dinstant lookups generally take 1-5 seconds. The five minutes is probabbly how long the name you requested stays active as valid for NAMED. If I use NCSA telnet or CUTCP I have a problem like you mentioned (for non local lookups). They do not seem to ask the name server again if it doesn't have the address ready right away. As to if it's the Apollo or the IBM that is at fault, I do not know... - John_Lauro@ub.cc.umich.edu University of Michigan - Flint
mheffron@orion.oac.uci.edu (Matt Heffron) (09/18/90)
In article <1990Sep17.134229.20117@caen.engin.umich.edu> jal@acc (John Lauro) writes: >In article <26F169EE.21884@orion.oac.uci.edu> mheffron@orion.oac.uci.edu (Matt Heffron) writes: >> >>I'm having trouble with running /etc/named under SR 10.2. >>It seems to run fine for a while, but if it is idle for very long >>(5-10 minutes, maybe longer,...) it then fails to respond within >>a reasonable time to a request, and the requestor times out (~30 sec) >>if the request is repeated immediately, then it works ok for awhile >>again. Does anybody have any ideas about this? >>Pertinant details: [ eliminated ] >>Matt Heffron mheffron@orion.oac.uci.edu > >I think at least part of the problem is with the software you are >running on the IBM. If it's something local that NAMED is an authority for, >I always get a quick response. The name server times out resolving the name when I do a ping TO THE NAMESERVER ITSELF!!! (pinging by name, the pointer to the 'default name server' is by IP number). The nameserver is an authority for about 8 hosts (**SMALL** /etc/named.hosts file.) and doesn't pass requests to any other name server (there aren't any here yet, and we aren't yet directly connected to the internet). > >As to if it's the Apollo or the IBM that is at fault, I do not know... > > - John_Lauro@ub.cc.umich.edu > University of Michigan - Flint -Matt Heffron mheffron@orion.oac.uci.edu