[2006] in Release_7.7_team
syncconf hack
daemon@ATHENA.MIT.EDU (Greg Hudson)
Sat Dec 4 16:19:55 1999
Date: Sat, 4 Dec 1999 16:19:30 -0500 (EST)
Message-Id: <199912042119.QAA20721@small-gods.mit.edu>
From: Greg Hudson <ghudson@MIT.EDU>
To: source-reviewers@mit.edu
CC: release-team@mit.edu
Background for release-team: hotline is noticing a lot of Ultra 5s and
10s (probably only ones with PUBLIC=true) losing because the four
network config files are showing up as zero length. Having looked at
such a machine, I believe this is because the machines have rebooted
uncleanly before the new config file data is written out.
hotline has asked for relief from this problem. As it is after drop
date, I think we want to do something very conservative. I suggest
applying this patch to syncconf and putting the new file out to field
early next week without bumping the patch level. PUBLIC=true machines
will take the new version of syncconf and hopefully stop seeing the
problem for the most part; PUBLIC=false machines are not as prone to
the problem because they don't re-sync the network configuration on
every boot.
Index: syncconf.sh
===================================================================
RCS file: /afs/dev.mit.edu/source/repository/packs/maint/os/solaris/syncconf.sh,v
retrieving revision 1.4
diff -c -r1.4 syncconf.sh
*** syncconf.sh 1999/11/10 06:01:12 1.4
--- syncconf.sh 1999/12/04 21:15:05
***************
*** 131,136 ****
--- 131,141 ----
append /etc/inet/netmasks "$ADDR $netmask"
fi
+ # Decrease the likelihood that the machine reboots
+ # uncleanly before the new files are written out to
+ # disk. A more robust algorithm is in the works.
+ sync
+
# Hostname configuration happens prior to rc2 scripts on
# Solaris.
if [ "$ADDR" = dhcp -a "$wasdhcp" = false ]; then
