On Sun, 8 Feb 1998 proff@suburbia.net wrote: Hmmmm. kill -1 on the master server caused it to exit. restarts are logging this info: Feb 9 00:27:58 news nntpcached[27043]: nntpcache.c:207:Invalid argument: SIGSEGV! Again, removal of cache.mmap allowed it to start. There was more than enough disk space this time (3Gb free, 4 million inodes avilable). > > 4: cache.mmap managed to be corrupted somewhow in the last couple > > of days. I assume it's because /var/spool/news/ ran out of space. > > It's possible. cache.mmap on create is that largest it will ever > be, however it is a *sparse* file to begin with. Inititially I thought this: Which is probably why it got corrupted. The expire trigger level was set to 10Mb, while the mmap was set to 128Mb. Last time I checked, the mmap was only actually using 12Mb. However, it may actually be being caused by innreport being run as root. I've changed the logfile rotation script to run that part as news, we'll know tomorrow if that's what caused it. > > 5: (related). If nntpcached starts an expire run and the master > > server dies, a new master server cannot be started until the expire > > process has finished or is killed. It took ~20 hours to clear out > > 5Gb space from my spool, which had users bitching loudly. > > The master server should never die :) killing an expire half way though > should be harmless. "should be" and "are" are 2 different things. :-) The master server here runs out of inittab, so if it's gone away, it's very dead. > > 6: spam.filter. Anyone got a working spam.filter? > > > > I had an extended filter I thought was OK, but 2.2 onwards refuse > > to start if it's in place. I've had to fall back to the distributed > > one. The non-working one is available for inspection and ridicule > > at ftp://news.manawatu.gen.nz/pub/news/spam.filter > > I can't seem access this esteemed ftp site. Can you tell me what the > complaints were about yoru old filters, post 2.2? Feb 9 00:34:55 news nntpcached[27261]: acc.c:233:Invalid argument: bad regular expression in spam.filter line 59: Inappropriate use of magic character at offset: 1 Here's what's in it (leading comment lines snipped): ##Subject: 100001 nocase (money|cash|dollar|\$.*\$|^\$|\$$)|(great|business).*(opportunity|venture)|(rich|succ?ess).*(fast|quick)|((free|f r e e) offer)|(spam ) From: 100001 nocase (*netwin.com.nz*|*sex.org|*netzilla.net) Message-ID: 100001 nocase (*fake@netwin.com.nz*) Newsgroups: 100001 nocase (*forsale*forsale*forsale*forsale*) Newsgroups: 100001 nocase (*.sex.*.sex.*.sex.*.sex.*.sex.*) article 10001 nocase *answerme.com*|*t-1net.com* If the Subject line is uncommented, the complaint changes to line 3. You should be able to reach ftp://news.manawatu.gen.nz. Then again I've been having problems getting to ftp.nntpcache.org the last few days, with traceroutes dying consistently at the world-net<->telstra gateway in Melbourne. They're now dying at this point: 18 mpxvic2.lnk.telstra.net (139.130.50.162) 522.444 ms 685.939 ms * 19 mel3.gw.mpx.net.au (198.142.142.4) 560.326 ms * * 20 * * * 21 * * * 22 * * * Trace times are high because we're getting from NZ to Australia via the USA. > Examine the web server and see what it says about the various log types > being on or off. (I just reviewed the log code and it all looks a-ok). As a FYI, you need a "Pragma: No-Cache" statement sent from the server or else proxy servers will cache the result (as mine has done) (For some reason port 9119 had died, which is why I hupped and discovered that the mmap file was corrupted again.) int umask 18 bool contentFilters true bool xoverFilters true bool logFromClient true bool logToClient false bool logFromServer false bool logToServer true bool logListMerge false bool logListMergeCorrelation false bool logNocem true bool logDebug false bool logInfo true bool logWarnings true bool logErrors true bool statistics true What's appearing in news.debug is lots of this: Feb 9 00:49:03 news nntpcache-client[27297]: sockets.c:427: <- ARTICLE 225488 Feb 9 00:49:03 news nntpcache-client[27297]: sockets.c:463: => [news2.triton.net.:1119] ARTICLE 225488 Feb 9 00:49:12 news nntpcache-client[27297]: sockets.c:427: <- ARTICLE 225490 Feb 9 00:49:12 news nntpcache-client[27297]: sockets.c:463: => [news2.triton.net.:1119] ARTICLE 225490 Also FYI: don't attempt nocem verification on a 486dx2/66. It won't keep up with the volume of nocems currently being posted. :-( My news server is now awaiting a new CPU and mobo. It was quite happily keeping up with 3-4Gb news a day as an INN box, but PGP verification of 500-700 messages per hour has suceeded where sheer bandwidth failed. Thankfully the puny 486 downstream boxes parked in local schools won't need nocem handling if they all feed off this one :-) AB
