[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: flexible restore/install system




On Wed, 19 Oct 2005, T wrote:

> ,-----
> | 
> | * No gold server. You work from the command line of any representative
> | target machine.

bad thing to have .... if it fails  and there is no silver or bronze
server with identical contents
 
> | * No central repository. Packages and change orders are stored in a
> | distributed cache, checksummed, replicated, and spread across all
> | participating machines.

repository is nice ...  if people need to be disciplined to comment
all their changes before releasing to the rest of the machines

	- many hundred ways to do that "task" so that only 
	tested files and patches gets out to the rest of the machines

> | * No CVS server. See the previous point. 

ditto

> | * No single point of failure. See above. 

ditto

> | * Better workflow.

if it takes more than a few minutes per day .. something is wrong

if the admin is afraid of 50 or 100 or 500 or 1000 or 5000 servers,
something is wrong withthe admin

the hardest part is to maintain and test the first 10-20 systems

and clone those patches/fixes onto the next 10-20 machines and
the rest of the world picks up its changes from the "release" servers

>  No more futzing around with CVS checkins, rsync updates,

that'd be a good or bad  thing ... depending on where you're looking

--------

for small world of machines ... say 5-25 ...

	make a cdrom with minimum drivers to support a network
	install off the net or clone any of your own local servers

	it'd be more (geeky) fun to boot from (network) floppy
	and do a network install  or even a usb-stick if you need
	more than 1.44/2.88MB to boot and install

	*
	* if you lose your cdrom or floppy or usb-stick,
	* you better have a few backup or a way to recreate a new one
	*

	- or do a network boot for all machines, so they are all 
	identical, except that the "pxe servers" will have to have
	a kernel that supports all the various client boxes

- you cannot get away from "single point of failure"

fun stuff and ez to do  ....  in 5 min or 5hr or 5 days ... would
depend on what else *one* needs to do for the other 100 hrs of the week
and what happens when any of the machine decides to go on holiday 
while you're on your honeymoon or vacation

c ya
alvin



Reply to: