I have autofs set up to work the same on three machines, i.e. i use the same autofs config files on all three. This means sometimes autofs is mapping via a link to a local directory, or sometimes actually mapping over the net. The problem has to do with when it maps via a link to a directory local to the machine. Sometimes (and I don't have any idea why it doesn't happen all the time), expiring doesn't appear to be successful. I get messages like this in my /var/log/messages: Nov 7 15:20:19 visar automount[17608]: expired /data/argon Nov 7 15:35:19 visar automount[17648]: expired /data/argon Nov 7 15:50:19 visar automount[17686]: expired /data/argon Nov 7 16:05:19 visar automount[17724]: expired /data/argon ad infinitum. the only fix seems to be to reboot the machine. I also notice messages when rebooting, AF_INET forgot to set XXXX when YYY. Fix it! etc. I am sure this is not the correct text, but it does not show up in log files, only on the console so it is a pain to capture the text of those messages. I am not sure that they have anything to do with autofs problem, but they appear to have to do with rpc which does have to do with autofs if I am correct.
Does this still happen with autofs-3.1.3-9 from Red Hat Linux 6.1 or autofs-3.1.4-2 from Raw Hide? Looking through the autofs code, you'd only see this happen if the removal of the link failed, because autofs appears to ignore the code returned when it attempts to remove the link.
The damndest thing ... I just checked my logs, and for some reason it is not happening anymore. Even stranger, ALL messages about expiring directories have vanished. Normally whenever I would log off and go home, about 15 min later there would be some message about expiration. Now there is nothing. On other machines, the log messages are still present (although I did not see any of the massive repetition which I posted the original bug report about.) The only things which are obviously changed is that on the machine which no longer has the problem (nor the normal expiry messages), I have modified inetd.conf so that linuxconf is commented out, and also on this machine I installed the "junkbuster" cookie-filtering program.
That is indeed odd. If the machines in question are all running the same versions of autofs and the kernel, they should all be behaving similarly wrt expirations. I can find no differences in this area between autofs 3.1.2 and 3.1.4, nor in the autofs filesystem driver in 2.2.12 and 2.2.14.
One more thing -- I went back thru my logs to find out when the autofs messages stopped. They stopped right after a reboot. I looked to see what had been changed in the month prior to reboot (this is RH 6.0) --- I must have installed a sysklogd patch since the rc.d file for syslog was changed. Perhaps this is why the logging changed? i cannot find much else. I have recently ordered RH 6.1 and will see what happens when I install it. Right now I have no clues (and also it is a non problem).
I doubt the syslog init script would have any effect -- it only starts and stops the service, and the configuration file (/etc/syslog.conf) is where the actual modifications would need to be made. I still get the occasional expire message in my log files here, but nothing repetitive. The NEWS file in /usr/doc/autofs* doesn't shed any light on this, either.
It''s been a while since the last entry on this bug. If there''s nothing else, I''ll go ahead and mark it as resolved.