[clug] Automated install and maintenance
Michael.James at csiro.au
Michael.James at csiro.au
Tue Jun 5 01:21:05 GMT 2007
On Mon, 4 Jun 2007 08:08:44 pm you wrote:
> That sounds a lot like System Imager
> [http://wiki.systemimager.org/index.php/Main_Page]
> any reason you didn't use it?
I had a look at something like it
but it looked a lot more static than real world situations need.
In their model of development I build the final master node,
image it and build the rest of the cluster from that.
Then if I want to use systemimager to change configuration
or update a package I have to re-build the image
and re-boot everything from the boot CD. Hassle
In my system the master node IS the image,
complete with all the changes I just did in vi.
If those changes are too widespread to be conveniently rsync-ed
then I fire a general reboot across the cluster.
So I can use cloneboot to ensure uniformity.
There is nothing special about the master node,
I can pick any node and un-enslave it
just by changing a link in /tftpboot.
Then I can develop it, re-partition its drive, upgrade the OS,
test and if all goes well; make it the new master.
Do a reboot of all but the old master and test again.
If there's a problem, just point the other nodes
back to the old master and reboot back.
Yes, once I blow the old master away, it's lost.
But I don't do that till the new config has been working for a good while.
It looked as if systemimager didn't do the partitioning.
With my system I could commission 6 new blades in 20 minutes,
that's from opening the crate to running with apps installed!
Cloneboot has been doing just what I want for 4 years now
and it's a few K of shell. Systemimager is 650Meg.
michaelj
--
Michael James michael.james at csiro.au
System Administrator voice: 02 6246 5040
CSIRO Bioinformatics Facility fax: 02 6246 5166
No matter how much you pay for software,
you always get less than you hoped.
Unless you pay nothing, then you get more.
More information about the linux
mailing list