Arran Cudbard-Bell wrote:
> Alan DeKok wrote:
>> Alan DeKok wrote:
>>>   I've just committed massive changes to the server core.  The "diff" is
>>> about 3k lines, and doesn't include deleted or added files.
>>   More code changes today:
>>   Multiple requests are proxied to a home server.  If the home server is
>> marked dead while the NAS is retransmitting the packets, the current
>> code (1.x) keeps sending the retransmissions to the dead home server.
>>   In the CVS head, it now discovers that the home server is dead, and
>> picks a live one from the appropriate server_pool.  When coupled with
>> the support for checking if a dead home server has come back to life,
>> the server should be MUCH more robust in the event of home server failure.
>>   i.e. With the current code, many proxied requests can get rejected,
>> even if there is a home server for the realm that is live.  With the new
>> code, all possible efforts are made to minimize the number of requests
>> that get rejected.
>>   No other server can do better than this.
>>   Alan DeKok.
>> --
>>       - The web site of the book
>> - The blog
>> - 
>> List info/subscribe/unsubscribe? See 
It seems to be occuring when the freeradius switches from the dead 
home_server to another in the pool.
Repeatable, happens on every transition to new proxy server.

Looks like a pretty generic memory allocation error, but I can provide 
platform / library information if it'll help.

When it works it'll be awesome :D

List info/subscribe/unsubscribe? See

Reply via email to