Hi,
The topic of HA is a touchy one, I worked for Stratus for a short while,
some of the mindset stuff sinks in fast. I would make the argument that
the first steps don't involve a lot of technology. I have repeatedly
seen businesses that have the resources take "the short route" and it
invariably bites them. The first step to good stability and availibility
is good process.
There are some "Best practices", there, the buzzword requirement is
done.
- Fix your processes and procedures
- Change management
- Test and development environments
- Well documented and planned deployments only
Then comes the basic technology parts.
- Always use a "real" server, this has the following attributes
- Hardware RAID
- Hot spare disk
- Redundant power supplies
- Major vendor built
- Hardware warranty with good turnaround
The mindset changes here will yield a fairly high gain.
After a business has these things nailed down they can start thinking
about real HA, fault tolerant hardware, clusters, etc. That is a game
that most folks don't have the dough to play. The ante is pretty high,
just to get in that game, and there is no end to how much money you can
spend.
Some day draw the 9's vs $s chart for your customer, after 5 9s the $
like becomes vertical. Most places can achieve ~99% uptime without any
duplication of hardware. That is aside from things like redundant disks
and power supplies.
/Soapbox.
In case you couldn't tell, my work life involves quite a bit of
explaining this kind of thing to a business that REALLY should know
better. ;')
Good luck to all your efforts to help customers get the most out of
their investment.
Kevin
> -----Original Message-----
> From: Greg Baker [mailto:[EMAIL PROTECTED]]
> Sent: Tuesday, July 09, 2002 8:14 PM
> To: [EMAIL PROTECTED]
> Subject: Re: [e-smith-devinfo] Some wish-list ideas for version 6.0
>
>
> On Wed, 2002-07-10 at 00:15, Richard Ford wrote:
> > High availability for SME's? Is this not seriously missing
> the mark on the
> > intended market segment? This is a 2x increase in costs
> where most SME's to
> > start with are hard pressed to pay for SCSI hardware raid
> and SCSI DAT/ETC
> > backup, ECC ram, etc.
>
> Sorta. Most of the customers I've seen so far don't go and shell out
> on any new hardware. They just use whatever junk is lying
> around which
> (a) is hardware that has already been paid for and/or depreciated,
> and so essentially free, and (b) very prone to failure.
> Doubling the
> cost of the hardware (twice nothing is nothing) isn't a big problem.
>
> I've got two quite small companies that are wanting high availability.
> One needs it because down-time costs them a lot; the other doesn't
> have anyone on site to fix anything -- the nearest technical person
> is several hours away. The former is relatively easy to fix -- they
> have two identical computers, one with removable hard disks, the other
> without. If something (hardware-wise) goes wrong they just
> put the disks
> into the other one. The latter is harder, (and I haven't
> done anything
> with them yet) and I'll probably end up using this How-To:
>
> http://www.e-smith.org/docs/howto/contrib/SME_High_Availabilit
> y_How-To.htm
>
> Which shows that there's other people out there wanting high
> availability SME servers.
>
> > Nice idea, but incorrect market, I would assume that most
> SME's would want
> > reliability and integrity before availability - which is a VERY big
> > umbrella.
> >
> > It's not like the SME server is used by FIFA.org,
> Olympics.com and the
> > international currency and futures traders?
>
> No, but a busy store (customer 1 above) loses a lot of money
> for every minute their computers are off-line. They actually did go
> and buy high quality hardware, etc. etc. but that's still not enough.
> I think they would have much preferred something automatic
> rather than
> manually moving disks around.
>
> > The marginal value to them would be lower than the price
> > of acquiring it.
> >
>
> Not always. It's not like I'm asking the Mitel people to go and
> write a h-a system. I'm just asking them for two lines of code that
> creates a hook to hang one on. If it never gets used, well, it's not
> a vast amount of R&D that has been wasted.
>
>
> > > > 2) I haven't asked Hsing Foo Wang about this, but it would be
> > > > nice to have some hooks for better integrating high
> availability.
> > > > I was thinking something like a hook like
> > > > if (-f /etc/ha-esmith/haconsole.pm) {
> > > > require 'haconsole.pm'; haconsole::do_hook();
> > > > }
> > > > somewhere around line 723 of /sbin/e-smith/console
> (where is asks for
> > > > server/gateway/both mode).
> > > >
> > > > This would make it easier to have "smart" high availability,
> > > > where it just asks for the IP address and root password for the
> > > > other high availability server, and then works out all the
> > > > remaining parameters (and sets up high availability
> automatically).
> > > >
> > > >
> > > > --------------
> > > >
> > > > Anyone have any thoughts or comments?
> > > >
> > > >
> > > >
>
>
--
Please report bugs to [EMAIL PROTECTED]
Please mail [EMAIL PROTECTED] (only) to discuss security issues
Support for registered customers and partners to [EMAIL PROTECTED]
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
Archives by mail and http://www.mail-archive.com/devinfo%40lists.e-smith.org
