At 09:01 PM 7/17/2003 -0700, you wrote: >Chris, > >We really don't give a damn about a fix, we just wanted Netwin to be aware of the >problem. True, if it perpetuates in future versions, we might revolt... But we can >either try 5.7c1 or (my preference), go back to the prior version which was working >just fine for us. > >Until debug logs can be provided (which could be a while because the > Channel closed s=11 >error has occurred only twice since upgrading to 5.7b1 and because my customer is >pressuring me to try 5.7c1 to see if that fixes this), here are my observations - for >what they are worth.
Very interesting, thanks. I'm not sure if I can do anything useful till I see that next log, I can't think of any significant changes to that code lately, I have a 'hunch' it's hitting some kind of odd response from upstream but I can't quite see why it should stop pulling for more than a few minutes at that point, although it may reset it's upto counter which might just about be the cause. Still I'm reluctant to make changes when I don't understand the bug yet. If you can live with that debugging setting and send me the log next time you spot it that would be most great thanks. re: debug log - no sorry there's no setting to get it to a seperate file. Thanks. ChrisP. >Check the code where the "s=11" comes from to see if the socket is getting >closed/reopened. I checked the firewall because nothing was getting out or in from >the dead suck feed. Since a > tellnews update_clear > tellnews exit >followed by > /usr/local/dnews/dnews_start.sh >immediately started a suck from the previously dead feed, the firewall wasn't the >problem - so the gauntlet from DNews to eth1 was impassible. > >The second time it happened, I did a > tcpdump -i eth1 -n | grep ip.ad.dr.es >(where ip.ad.dr.es is the ip of the dead suck feed) >and there were no packets. > >So I set loglvl to info and then ran a > tellnews updateone alt.os.linux.slackware 3 >and (shame on me I didn't save the log) the log indicated increasing timeouts in the >29000 range and higher, but the (still running) tcpdump STILL showed no packets - for >as long as I had patience to wait (about 2 minutes, probably). > >(Turns out that updateone may not have been a good idea <shrug>.) > >So I repeated the > update_clear > exit > dnews_start.sh >and in less than 30 seconds I had a screenful of (still running) tcpdump output. > >While on the subject of logs and output, is it possible to redirect the debug output >to a file other than dnews.log? I _REALLY_ would rather not garbage up days/weeks of >logs with debug output (even if only suck output) if it can be prevented! > >I want to take this opportunity to shout: > THERE IS NO BETTER NNTP PROGRAM THAN DNEWS! > THERE IS NO BETTER SUPPORT FOR ANY COMMERCIAL LINUX PROGRAM THAN AT NETWIN! > THANK YOU FOR BOTH OF THE ABOVE! > >But on a personal note, I gotta say, "I'm glad HE paid for it, not me!" <grin>. > >buckprivatemil
