The good news is the BIND DNS server now initializes to the point of being
ready to accept queries.  The bad news is that it crashes every time right
after that point.

After dumping and restarting the machine, I looked at the POSIX bytestream log
file (created by Syslog/iX) that records BIND events.  I had caused several
records to be written to the log before the crash.  I had even displayed
these records before the crash with "tail".  But after the crash these
records (along with just about all log records from current syslog job) had
disappeared from the end of the file!

So I modified Syslog/iX to FSETMODE(,1) on the log file, and I re-crashed the
machine.  The FSETMODE attaches the log file to XM, and after the crash, XM
had managed to do a better job of preserving the end of the log file, but still
some records were lost.  I suppose it's possible that some of these records
were lost in the networking UDP layers and hadn't reached syslogd to be written
to the disk log file.  Not much I can do about that....

To make post-crash integrity even better for the records that do get written
to disk, do I also need to FCONTROL(,2) and/or FCONTROL(,6) after every disk
write by syslogd?
--
Mark Bixby                      E-mail: [log in to unmask]
Coast Community College Dist.   Web: http://www.cccd.edu/~markb/
District Information Services   1370 Adams Ave, Costa Mesa, CA, USA 92626-5429
Technical Support               +1 714 438-4647
"You can tune a file system, but you can't tune a fish." - tunefs(1M)