[RADIATOR] Missing info from error message

Jethro R Binks jethro.binks at strath.ac.uk
Tue Nov 26 09:14:42 CST 2013


On Tue, 26 Nov 2013, Heikki Vatiainen wrote:

> On 11/26/2013 10:55 AM, Jethro R Binks wrote:
> > Running 4.12.1, I get:
> > 
> > WARNING: ProxyAlgorithm EAPBALANCE declines to break up an EAP stream 
> > after detecting failure of Host ::
> > 
> > I see from other posts and the source there is supposed to be host and 
> > port information between those colons, but it doesn't seem I get it.  I 
> > can't think that my config can influence this.  Has a bug crept in 
> > somewhere?
> 
> That part of code has been unchanged for a long time. I can see the 
> above happening when you receive a request which has State attribute 
> with value such as EAPBALANCE:id=n where n is not a number or equal to 
> or larger than the number of Hosts you have configured with AuthBy 
> EAPBALANCE.
> 
> For example, if you have 3 Hosts configured, then 0 <= n < 3
> 
> How often do you see the error message?

Not very often, today:

Tue Nov 26 08:35:42 2013: WARNING: ProxyAlgorithm EAPBALANCE declines to break up an EAP stream after detecting failure of Host ::
Tue Nov 26 08:35:49 2013: WARNING: ProxyAlgorithm EAPBALANCE declines to break up an EAP stream after detecting failure of Host ::
Tue Nov 26 08:35:57 2013: WARNING: ProxyAlgorithm EAPBALANCE declines to break up an EAP stream after detecting failure of Host ::
Tue Nov 26 08:36:05 2013: WARNING: ProxyAlgorithm EAPBALANCE declines to break up an EAP stream after detecting failure of Host ::
Tue Nov 26 11:01:45 2013: WARNING: ProxyAlgorithm EAPBALANCE declines to break up an EAP stream after detecting failure of Host ::
Tue Nov 26 11:01:53 2013: WARNING: ProxyAlgorithm EAPBALANCE declines to break up an EAP stream after detecting failure of Host ::
Tue Nov 26 11:02:01 2013: WARNING: ProxyAlgorithm EAPBALANCE declines to break up an EAP stream after detecting failure of Host ::
Tue Nov 26 11:02:09 2013: WARNING: ProxyAlgorithm EAPBALANCE declines to break up an EAP stream after detecting failure of Host ::

But some days there are up to 40 or so of them, in clusters ~8 seconds 
apart within a cluster.

> > "ProxyAlgorithm EAPBALANCE declines to break up an EAP stream after 
> > detecting failure of Host 
> > $host->{Name}:$host->{AuthPort}:$host->{AcctPort}", $p);
> 
> The code is careful to not use $host when it is not defined, but the
> error message does not care. Are you using multiple EAPBALANCE systems
> where the other systems may have a larger number of Hosts configured?

I'm only using EAPBALANCE in one context:

## For authentication of visitors to here using eduroam
<AuthBy EAPBALANCE>
        Identifier      EduroamVisitor
        AuthPort        1812
        AcctPort        1813
        FailureBackoffTime      10
        RetryTimeout            5
        Retries                 1
        AutoMPPEKeys
        UseExtendedIds
        # We do not want to see these coming from offsite:
        # We only want to send these attributes with eduroam authentications:
        AllowInReply    User-Name,Reply-Message,State,Class,Message-Authenticator,Calling-Station-ID,Proxy-State,EAP-Message,MS-MPPE-Send-Key,MS-MPPE-Recv-Key,Acct-Status-Type,Acct-Session-ID,Operator-Name,Chargeable-User-Identity,Framed-MTU
        <Host roaming0.ja.net>
                Include         %D/secret.%{GlobalVar:thishost}.roaming0.conf
        </Host>
        <Host roaming1.ja.net>
                Include         %D/secret.%{GlobalVar:thishost}.roaming1.conf
        </Host>
        <Host roaming2.ja.net>
                Include         %D/secret.%{GlobalVar:thishost}.roaming2.conf
        </Host>
</AuthBy>

where roamingX.ja.net is our National Proxies (hi Alan). (I haven't 
touched this bit of config in years).

However I am looking at breaking out my config into a front-end and 
several backends to spread load across cores where I'd be using 
EAPBALANCE, which is why my eye was drawn to it as I was thinking about 
it.

> If the problem persists, can you switch on debugging to see what triggers
> the above.

Busy server and what looks like a reasonably uncomon event, 
could be tricky!

Is this short thread relevant I wonder:

  http://www.open.com.au/pipermail/radiator/2013-May/019100.html

Thanks,

Jethro.

.  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .
Jethro R Binks, Network Manager,
Information Services Directorate, University Of Strathclyde, Glasgow, UK

The University of Strathclyde is a charitable body, registered in
Scotland, number SC015263.


More information about the radiator mailing list