Jeff Ferland's questions

I've tried multiple pieces of DHCP client and software in attempting to setup a solution for allocating a /127 prefix to virtual machines so that each maintains its own layer 2 isolation. Because there would only be one host assigned to each network, a /64 is impractical.

While the prefix size could reasonably be somewhere in the /64-127 range, the crux of the problem has been the same regardless of the software used in configuring: the DHCP call to bring up the interface uses the address advertised by DHCPv6 and inserts two routes: the /127 given by the router advertising packets and a /64 as well.

Any thoughts on why I'm getting the additional route added across dhcp client vendors?

When running ls -dZ /, I get the following output:

system_u:object_r:root_t:s0

Can the user and role fields have any effect access permissions involving a file? Do they mean anything?

The following failure occurs since /dev/shm is a symlink to /run/shm on newer Debian and Ubuntu systems. Is there a way to adjust this behavior in Kickstart so it tries to also mount from /run/shm? I would find this preferable to changing my default symlink to a bind mount. Ideally, this would be something I could define in the .ks file.

chroot builddir/media/centos6/x86_64/tools livecd-creator --config /build/livecd.ks

## Snip -- output from mke2fs and tune2fs

mount: special device /dev/shm does not exist
Error creating Live CD : Bind-mounting '/dev/shm' to '/var/tmp/imgcreate-ubTScl/install_root//dev/shm' failed
make: *** [builddir/media/centos6/x86_64/image.iso] Error 1
user@host:/tmp/workspace$ ls -lh /dev/shm
lrwxrwxrwx 1 root root 8 Aug 24 14:21 /dev/shm -> /run/shm

I have an image in my head where a link is too slow to realize the real-time transfer of files, but fast enough to catch up every day. What I'd like to see is a master <-> master setup where when I write a file to Server A, the metadata will transfer to Server B immediately and the file will transfer at idle or immediately when Server B's client tries to read the file before Server A has sent it.

It seems that there are many filesystems which can perform well over fast links, but I don't know of any that do well with a big bottle neck and a few hours of latency.

Upon cloning a base image, I need to reconfigure basic settings. Regenerating the ssh host key, changing static IP assignments, setting the host name, etc.

Because of the network setup, DHCP is not an option. That more or less rules out SSHing in with a predefined key or running a startup script since I can't provide the IP externally.

I'd most like to mount the filesystem of the new machine on Dom0, but the lvm volumes are exported and it appears to be Bad Form to import them so the Dom0 machine can see them.

What's your best suggestion for altering files in a cloned VM before boot? Must be non-interactive, and I'm going to guess out the gate that scripting access via xe console is not going to work well.

Scenario: my configuration file is defined by a .erb file which includes the snippet below.

<% backupclients.each do |backup_files| -%>
Job {
  Name = "Server"
  JobDefs = "DefaultJob"
  Client = <%= backup_files %>-fd
  Pool = TeraMonth
  Write Bootstrap = "/var/lib/bacula/<%= backup_files %>.bsr"
}
<% end -%>

The server's configuration file needs a repeated entry for each client host. If I were to create a simple array, this would work without issue. What I want to do, however, is have each host register itself and then collect the data using the <<| |>> pragma similar to what one would do with the nagios_* types.

The standard example for this involves exporting a type.

class ssh {
  @@sshkey { $hostname: type => dsa, key => $sshdsakey }
  Sshkey <<| |>>
}

However, I can't quite figure out how to write a type or reference it in a way that would allow me to read that array of values from the .erb template. Is there a way I can use exported resources in combination with a variable loop in a .erb file?

I'm setting up my Bacula configuration in Puppet. One thing I want to do is ensure that each password field is different. My current thought is to hash the hostname with a secret value that would ensure each file daemon has a unique password and that password can be written to both the director configuration and the file server. I definitely don't want to use one universal password as that would permit anybody who might compromise one machine to get access to any machine through Bacula.

Is there another way to do this other than using a hash function to generate the passwords?

Clarification:

This is NOT about user accounts for services. This is about the authentication tokens (to use another term) in the client / server files. Example snippet:

Director {                            # define myself
  Name = <%= hostname $>-dir
  QueryFile = "/etc/bacula/scripts/query.sql"
  WorkingDirectory = "/var/lib/bacula"
  PidDirectory = "/var/run/bacula"
  Maximum Concurrent Jobs = 3
  Password = "<%= somePasswordFunction =>"         # Console password
  Messages = Daemon
}

CTRL+A, for example, is printing "^A" instead of returning my cursor to the start of the line.

This happens via ssh, in a screen session, or on the hard console. $TERM shows either xterm or linux depending on which one I'm on.

How do I get the keymap fixed?

I have a file that was deleted, but is still held open by a program. I found the inode number using lsof. How can I create a hard link back to that inode?