Re: VAX hardware support alternatives
From: brandon@dalsemi.com
Date: 04/12/03
- Next message: Don Sykes: "Re: Reject Unknown Users from SMTP"
- Previous message: Larry Kilgallen: "Re: HP, P&G, Ericsson"
- Maybe in reply to: Rich Jordan: "VAX hardware support alternatives"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ] [ attachment ]
From: brandon@dalsemi.com Date: Fri, 11 Apr 2003 18:08:42 -0500
> So run them in the cluster with the live machines. Fairly easy to see if any
> machine, live or spare goes bad, and you can then even look at automatic
> failover.
My thought exactly. An excellent thought at that! ;-)
I have had spare disk drives sit on the shelf for less than a year and when I
put them into a Storage Array they failed within days. These were fairly new
drives and had been used for a few months before being retired...
As you say - add them to the existing cluster (or create a cluster) mixing both
online and spare machines in that cluster.
Concerns:
1) License:
If you will be allowing users to access the spare servers then you
will need to have additional licenses. If you will only allow users to access
the spare servers in the event of a failure then no real issue here.
Be aware of keeping the servers similiar - if the applications are CPU or server
specific you will have problems.
2) Failover:
Set up an automated failover procedure. No need to have a failover server if
it is not automated - in my opinion.
Create a heart-beat mechanism - using DECnet or TCP/IP. This monitors the
production server and detects and responds to failures.
Then test it! Setup a schedule to do this. First of the month, every quarter,
but schedule it! Too many times we find out our failover server did not work
because we changed something on the production side - and oops, I did it again.
3) DNS:
Some DNS allow for load balancing and auto-check of existing service. If #1
(licenses) are not an issue, load balance the production with the spare
servers. Weight the production server heavier than the spare - or you could
balance it out. Have DNS serve only available IP addresses.
Use CLUSTER alias too.
4) Server characteristics:
You might want to keep the servers the same. You could drop down to a
lower-end server; however this would run your environment at a reduced rate.
Sometimes this is acceptable.
5) Storage Array:
Make sure both production and spare servers have direct access to the storage.
Does no good to have a production server with local disk attached and the spare
is unable to gain access to them.
John Brandon
VMS Systems Administrator
Dallas Semiconductor
john.brandon@dalsemi.com
972.371.4172 wk
972.371.4003 fx
- Next message: Don Sykes: "Re: Reject Unknown Users from SMTP"
- Previous message: Larry Kilgallen: "Re: HP, P&G, Ericsson"
- Maybe in reply to: Rich Jordan: "VAX hardware support alternatives"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ] [ attachment ]
Relevant Pages
|
