John,
It did end up being something that Bastille did to the system in tightening
it up. After beating my head against the fortress walls for a week and not
being able to get it to run the script, the Truss, nor the amandad itself at
all, we finally determined that it had closed down anything that used UDP
and dgrams.
I finally bit the bullet tonight and rebuilt the main server from scratch
(Amanda and Bastille was all that I had configured so far). This time when I
ran Bastille, I made two changes: 1) I told it not to take strong measures
against the BSD r-tools, and 2) Don't add the TMP-DIR scripts. One of the
two of these fixed the problem, but I am not sure which. Given that it takes
a couple of hours to rebuild (an that it is working now) I am not sure I
want to find out right now.
I just now finished my initial testing and have it amchecking against the
local host and one remote RH7.1 client, and am about to embark on the
compiling for RH6.2 and Solaris 7/8. Then I have to give serious thought to
how I really want it to rotate through these 20 tapes... Too bad there isn't
one big button that says "Backup Everything" that figures all this out for
you ;-).
>From what I have seen so far this is a great program. Keep up the good work!
markh
-----Original Message-----
From: John R. Jackson [mailto:[EMAIL PROTECTED]]
Sent: Friday, August 31, 2001 10:08 PM
To: Mark Holm
Cc: '[EMAIL PROTECTED]'
Subject: Re: Can't start amandad from xinetd on RedHat 7.1
>... I have gone through the archives and
>tried all of the steps suggested in all of the archived articles on this
>subject without success. ...
The outputs you posted look reasonable.
>... other than that the amandad
>service was looping and xinetd was killing it. ...
OK, then it's something about the way xinetd is starting amandad
that's bad. And it's happening right away before amandad can get even
barely started.
One possibility is that you should probably add "groups = yes" to the
xinetd amandad config file so it gets started with the alternate groups.
The default is to only let it have its primary group.
I'd suggest going back to the script and truss tests. One possiblity for
why they didn't work before is that after xinetd terminates a service, it
leaves it dead for several minutes and isn't even listening. That would
look the same from the amcheck side -- host down -- but for a totally
different reason. Before each and every test, do a refresh of xinetd
and "netstat -a | grep amanda" to be sure someone is listening.
>Mark A. Holm
John R. Jackson, Technical Software Specialist, [EMAIL PROTECTED]
