datpete@daimi.dk (Peter Andersen) (07/05/90)
We run C-news patchlevel 16-Apr-1990 with dbz and nntp 1.5.7, and have experienced the following problem: Due to a major system-crash some control files were damaged. Concerning the news system this was the case for either history.dir or history.pag or both. However I took quite a while to figure out that this was the case. I did not get any mails or syslogs reporting troubles from expire, but I could see that it did not succeed, and dumped core in the NEWSCTL directory. If I tried running doexpire manually I got the following message: /var/spool/news/lib/newsbin/expire/doexpire: 19606 Memory fault - core dumped I discovered that it was the dbz-files that was corrupted when I saw that nntpd reported things like Jul 2 19:47:17 daimi nntpd[2243]: openartbyid: dbminit /var/spool/news/lib/news/history: Operation would block and Jul 2 19:47:19 daimi nntpd[2243]: openartbyid: dbminit /var/spool/news/lib/news/history: Not owner and Jul 2 19:47:38 daimi nntpd[2243]: openartbyid: dbminit /var/spool/news/lib/news/history: No such file or directory. (and I got a lot of mails from nntpd mentioning bad input batches) The history file was indeed present and was owned by news, so the two last messages was a bit confusing. But since examining the core file dumped by expire with adb showed that expire also had problems with dbminit, I realized that the dbz-files were corrupted and did a mkdbm which solved the problem. My question is: In most other cases the various C-news programs is rather verbose and mails all potential problems to me. So shouldn't it be able to realize that the dbz-files are corrupted, and tell me that. It seems that it *is* possible since nntpd is able to. In my particular case such a message would have spared me several hours of work to figure out what was wrong. Is this a bug in expire, or can I have a wrong configuration ? Peter Andersen
henry@zoo.toronto.edu (Henry Spencer) (07/06/90)
In article <1990Jul4.212921.26279@daimi.dk> datpete@daimi.dk writes: >Due to a major system-crash some control files were damaged... >... I realized that the dbz-files were corrupted >and did a mkdbm which solved the problem. > >My question is: In most other cases the various C-news programs is rather >verbose and mails all potential problems to me. So shouldn't it be able to >realize that the dbz-files are corrupted, and tell me that... I would have expected it to do so, in fact. Normally it is quite picky about the format of the files. I would appreciate more details, if you have any. (In particular, knowing precisely which function dumped core would be helpful.) -- "Either NFS must be scrapped or NFS | Henry Spencer at U of Toronto Zoology must be changed." -John K. Ousterhout | henry@zoo.toronto.edu utzoo!henry