Updating a mesh network full of machines.

[James Cameron]

On Wed, Jun 04, 2008 at 05:52:33PM -0400, C. Scott Ananian wrote:
> Other folks have reported success using screen or Konsole to send
> identical commands over ssh to a large number of machines, which is a
> lower-barrier-to-entry option.

Specifically the success I've had with a test bed of eight units was
with distributed shell, aka
http://www.netfort.gr.jp/~dancer/software/dsh.html debian package dsh
... either preload the image with keys for a remote accessor, or add a
key to each newly reimaged laptop.

To reboot the entire test bed ... "dsh --group xo reboot"

I've also used a time synchronisation script in earlier builds, might no
longer be necessary ... by placing an rdate binary in /usr/local/bin and
then ...

dsh --group xo --concurrent-shell "/usr/local/bin/rdate 10.0.0.1 && hwclock --systohc && cp /usr/share/zoneinfo/Australia/NSW /etc/localtime && date --rfc-3339=ns"

The distribution of the output is then useful to determine how well the
time was set across the test bed.  Running it a second time gives better
synchronisation, since everything is in memory.

-- 
James Cameron    mailto:quozl at us.netrek.org     http://quozl.netrek.org/