[Outages-discussion] [outages] UK wide DSL broadband outages

[Frank Bulk]

Thanks for explaining.  Sounds like they need to scale their RADIUS
infrastructure and also randomize that one hour interval.

Frank

-----Original Message-----
From: outages-discussion-bounces at outages.org
[mailto:outages-discussion-bounces at outages.org] On Behalf Of Paul Lettington
Sent: Friday, October 21, 2011 9:06 AM
To: outages-discussion at outages.org
Subject: Re: [Outages-discussion] [outages] UK wide DSL broadband outages

On Fri, Oct 21 2011 08:10:02 -0500, Frank Bulk wrote:
> Why did a 15-minute outage cause subtended customers to be out for much
> longer than that?

Because BT's RADIUS servers which look at the realm of the login and route
the
session to the correct ISP are unable to cope with the load of all the ADSL
users in the country continually reconnecting during an outage like this.

So when BT can't connect the end user to the ISP they accept the end user's
session themselves with a RFC1918 IP address for exactly an hour and not
connected to the internet, to stop them trying to log in.

In today's outage, even though connectivity from BT to the ISPs had returned
within 15 minutes, users who didn't reboot their routers or force a new
login
were stuck on the holding sessions until the hour expired. Also the sudden
load
of reconnections exactly an hour after the outage started meant that BT's
RADIUS servers still couldn't cope, and put most of the customers back on a
holding session for a second hour.

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