[18964] in Athena Bugs
Re: athena 9.0 (current almost-beta) install bugs
daemon@ATHENA.MIT.EDU (Garry Zacheiss)
Thu May 17 23:12:06 2001
Message-Id: <200105180312.XAA06426@dale.mit.edu>
To: lcs@MIT.EDU
cc: bugs@MIT.EDU, miki@MIT.EDU
In-Reply-To: Your message of "Thu, 17 May 2001 14:33:33 EDT."
<CMM.0.90.4.990124413.lcs@defiant.mit.edu>
Date: Thu, 17 May 2001 23:12:01 -0400
From: Garry Zacheiss <zacheiss@MIT.EDU>
>> Here's an odd problem in the install (current as of 16-may 17:00) on
>> solaris that precipitated some other problems:
Something you didn't specify in your mail that's relevant is that the
install was done off of the test install server (sunburn); the miniroot
there is configured to use installation scripts that aren't checked in
anyplace else and have some problems, unfortunately.
Considering that this is the installer for the release currently in
beta and not production software, testers@mit.edu might have been a
better choice than bugs.
The problem was that /sbin/athenainstall (called from /sbin/startup,
which is run by init) was failing to populate /etc/myIPaddr when this
code was run:
/sbin/hostinfo $hostname | grep 18 | awk '{print $3}' >/etc/myIPaddr
This was because named wasn't running, so hostinfo wasn't able to
resolve the hostname. The line above is also *wrong* because it assumes
that you're on net 18, which isn't a valid assumption to make. Miki,
this needs to be fixed.
/sbin/rcS had code in it that looked like this:
echo "starting named"
/etc/athena/named >/dev/null &
which was failing to start named for some reason; I changed it to:
echo "starting named"
/etc/athena/named
(named daemonizes itself) and named then started and the script was able
to populate /etc/myIPaddr correctly.
But wait, there's more. The installer would also fail if you didn't
specify a custom install, because the default would link /srvd and /os
to places that don't exist (there aren't 9.0 system packs in the athena
cell yet). I fixed install1.sh so a default install would work; this
can be reverted once we go into early test and have packs in the athena
cell for the new release.
Lastly, I noticed that the version of install1.sh in sunburn's
miniroot is based on an out of date version that still partitions disks
into /, /usr, and /var; it does not contain my changes to make all disks
>4GB be partitioned using the new partitioning scheme. These should be
merged in.
I haven't done anything to make syncconf more robust in the
presence of empty variables.
Garry
