[prev in list] [next in list] [prev in thread] [next in thread] 

List:       outages-discussion
Subject:    Re: [Outages-discussion] Avoiding puppet/cfengine as the next SPOF
From:       "Patrick W. Gilmore" <patrick () ianai ! net>
Date:       2014-01-27 17:34:54
Message-ID: B88876AE-FC49-4753-B39E-BA7115C098B5 () ianai ! net
[Download RAW message or body]


On Jan 25, 2014, at 20:38 , George William Herbert <george.herbert@gmail.com> wrote:

> Even with one-few-many, you have test failures, config problems that take longer \
> than test time to fully manifest, n^2 and n! Scaling for some disturbance patterns, \
> etc. 
> Nothing is perfect...

What he said.

Anyone who thinks you can manage a "planetary scale" (as we liked to say at Akamai) \
deployment without hitting a few hiccoughs is deluding themselves. As with pretty \
much everything, there is no single solution, just strategies to lower the risk. \
"One-few-many" is a strategy, and a good one, but not a solution.

However, I would still much rather be in the regime of "it is my fault we went down", \
than "someone else can take us down". Having recently transitioned from A to B, I \
feel this difference quite acutely and can tell you from personal experience, A is \
WAAAAAAAAAAAAY better than B.

-- 
TTFN,
patrick


> On Jan 25, 2014, at 4:48 PM, Jay Ashworth <jra@baylink.com> wrote:
> 
> > [ moved to -discuss ]
> > 
> > ----- Original Message -----
> > > From: "Conrad Heiney" <conrad@fringehead.org>
> > 
> > > Centralized configuration is necessary. It is also the new SPOF.
> > 
> > I would assert that if it is, you're not driving it correctly. 
> > 
> > Sure, you need it on things as big as the Googleplex.  But that doesn't mean
> > that you can't use Tom Limoncelli's celebrated "one, few, many, all" 
> > staged deployment approach, when setting up pushes.
> > 
> > In fact, it's *easier*, since you know that they'll all get the same 
> > configuration.
> > 
> > You have to be careful about interdependencies, of course, but that --
> > again -- is something computers are pretty good at.  :-)
> > 
> > Cheers,
> > -- jra
> > -- 
> > Jay R. Ashworth                  Baylink                       jra@baylink.com
> > Designer                     The Things I Think                       RFC 2100
> > Ashworth & Associates       http://www.bcp38.info          2000 Land Rover DII
> > St Petersburg FL USA      BCP38: Ask For It By Name!           +1 727 647 1274
> > _______________________________________________
> > Outages-discussion mailing list
> > Outages-discussion@outages.org
> > https://puck.nether.net/mailman/listinfo/outages-discussion
> 
> _______________________________________________
> Outages-discussion mailing list
> Outages-discussion@outages.org
> https://puck.nether.net/mailman/listinfo/outages-discussion
> 


_______________________________________________
Outages-discussion mailing list
Outages-discussion@outages.org
https://puck.nether.net/mailman/listinfo/outages-discussion


[prev in list] [next in list] [prev in thread] [next in thread]