msc@qubix.UUCP (Mark Callow) (11/09/84)
Every week we take our 11/750 running 4.2BSD down to single-user for
backups. Almost every week the rlogin daemon refuses to run upon
return to multi-user operations. After recompiling with -DDEBUG
to persuade it to print some error messages, rlogind says
"bind: address already in use"
Obviously the previous rlogind didn't die cleanly when the system went
to single-user.
[Minor Flame]
Why does rlogind swallow all error messages. I'd certainly like to get
messages when it has problems. It doesn't even print anything on the
console or via syslog. Is this a "feature" or a bug??
[Extinguish]
Does anyone know why rlogind didn't die cleanly? The backups are done
early in the morning and as afar as I know there are no active rlogin
connections at the time we go to single-user. However there are
undoubtedly some in state FIN_WAIT_2 due to the infamous bug.
We have no problems with any of the other network daemons.
What I can do about it? I'd like to get rid of the problem, but I'd settle
for a means of getting rlogind back to life without rebooting the system.
--
From the TARDIS of Mark Callow
msc@qubix.UUCP, qubix!msc@decwrl.ARPA
...{decvax,ucbvax}!decwrl!qubix!msc, ...{amd,ihnp4,ittvax}!qubix!mscrichl@daemon.UUCP (Rick Lindsley) (11/13/84)
Those FIN_WAIT_2's (or any other 'hung' network connection) is the reason
you cannot restart the daemon. You can use adb to zap those connections
(the technique was posted a while back) but since you just went down to
single user for backups anyway, why not reboot?
Rick Lindsley
...{allegra,ihnp4,decvax}!tektronix!richl
richl@tektronix.csnetloverso@sunybcs.UUCP (John Robert LoVerso) (11/15/84)
> Every week we take our 11/750 running 4.2BSD down to single-user for > backups. Almost every week the rlogin daemon refuses to run upon > return to multi-user operations. > Does anyone know why rlogind didn't die cleanly? The backups are done > early in the morning and as afar as I know there are no active rlogin > connections at the time we go to single-user. However there are > undoubtedly some in state FIN_WAIT_2 due to the infamous bug. > > We have no problems with any of the other network daemons. We had the same problem on three of our systems running 4.2 (a 780 and two 750s). After two to three days, rlogind would refuse to run anymore. Not having the time to track this down, we just agreed upon to have a full reboot done after dumps were completed. If there is a REAL fix, I'd like to know... John -- John Robert LoVerso @ SUNY Buffalo (716-636-3004) LoVerso%Buffalo@CSNET-RELAY -or- ..!{watmath|rocksanne}!sunybcs!loverso