Debian
sysvinit package

  1. Bug #275451
  2. Comment #27

Comment 27 for bug 275451

Revision history for this message
EmonkIA (jason-woods-deactivatedaccount) wrote :

DISTRIB_ID=Ubuntu
DISTRIB_RELEASE=9.04
DISTRIB_CODENAME=jaunty
DISTRIB_DESCRIPTION="Ubuntu 9.04"

Linux rambc 2.6.28-14-generic #47-Ubuntu SMP x86_64 GNU/Linux

Terminal server using Ubuntu Desktop install

I have found this bug to exist on a server that was upgraded from 7.04 directly to 9.04 (in steps). The problem was noticed after the upgrades were finished (the system was barely used in between upgrades). System is very customized as it is a terminal server. I have a LUN copy of the original 7.04 machine running on a blade if comparison is needed.

The problem appears to possibly be DNS related, but is definitely a race at boot. I use /etc/network/interfaces to hard code the static IP to the network interface (NetworkManager not installed). The NFS mount points were and still are in /etc/fstab, exactly has they have been (no change in upgrade). The NFS mounts use the server name via DNS to mount. The /proc/mounts does not show traces of "half mounted" filesystems.

If I put the NFS server name in /etc/hosts, the filesystems will automatically mount.
If I remove the NFS server name from /etc/hosts, the filesystems will not mount.
Adding "mount -a" in /etc/rc.local does not work.
Adding individual mount commands in /etc/rc.local does not work.

If I wait until the system is fully booted, "mount -a" will work.

Lines from /etc/fstab are pretty normal (all on one line, broken for ease of sight):
nas-03:/u64dev-s
  /home/shared nfs rw,rsize=8192,wsize=8192,hard,intr,timeo=14,bg 0 0
nas-03:/PSExport-ACCT
 /home/shared/PSExport/ACCT nfs rw,rsize=8192,wsize=8192,hard,intr,timeo=14,bg 0 0
nas-03:/PSExport-HR
 /home/shared/PSExport/HR nfs rw,rsize=8192,wsize=8192,hard,intr,timeo=14,bg 0 0

Line required in /etc/hosts contains an IP for "nas-03". This kludge is highly not preferred.

Lines from /etc/network/interfaces (munged):
auto lo
iface lo inet loopback

auto eth0
iface eth0 inet static
address 10.11.12.20
netmask 255.255.255.0
gateway 10.11.12.1

This bug will very soon be affecting us on another 17 servers, all being upgraded from v6.06 and v7.04 to v9.04. Hard coding multiple DNS entries will cause massive loss of hair and years off life. Not now, and possibly not for me, but someone at some random point in the future.