On Saturday 23 September 2006 18:53, Ian Cheong wrote:
> We had this discussion on failure tolerant clusters a while back - at
> the time, nobody owned up to having one.

And it might be expensive overkill.

While I am constantly experimenting with automatic failover recovery through a 
replicated server, so far our "tried and proven" simple & cheap solution has 
not disappointed us:

1.) All relevant software and data is hosted on a single computer, 
the "server".

2.) All people interacting with that software / data are connected 
via "clients" - either thin or fat, but only requirement to a client computer 
is the ability to run the X protocol and ssh. Meaning I can take a brand new 
laptop with a blnk harddisk, boot from a Knoppix CD, and can start working as 
soon as the CD has booted

3.) The server does not depend on any internal harddisks - it is connected via 
eSATA to an external RAID cage. The hard disks in the RAID cageare hot 
swappable. Apart from sending an email to the administrator(s), the RAID cage 
emits an audible alarm and displays error messages on a LED screen if one 
harddisk fails

4.) All our "client" computers are equipped with eSATA connectors and can act 
as instant drop-in servers in case the server fails

5.) We have a second spare RAID cage in case the RAID cage fails

6.) Offsite backup via rsync of server to hospital and my house twice daily 
(in case the surgery burns down or some such)

7.)  all computers are connected to good UPS

8.) spares for network cables, keyboards, mice, switches and ADSL modem on 
site + notebook that can replace either server or client.

If you do the numbers you will see that our solution is actually CHEAPER than 
having fully installed client computers for everybody, it requires only a 
fraction of the administrative effort, and yet it is far more failproof.

Horst
_______________________________________________
Gpcg_talk mailing list
[email protected]
http://ozdocit.org/cgi-bin/mailman/listinfo/gpcg_talk

Reply via email to