On Sep 01, 2010 at 14:50, Alex Balashov <abalashov(a)evaristesys.com> wrote:
On 09/01/2010 01:25 PM, Andrei Pelinescu-Onciul
wrote:
The problem with having fallback to the system
resolver is that in case
the entry is not in /etc/hosts (or similar), you will have double
queries for each failure and you would also not be able to do any dns
fallback / srv rec. load balancing (which depend on having cached
records).
There is no SRV resolution in the stub resolver?
There is (both SRV and NAPTR), but there is no way to fallback
(so if the first destination fails it won't try another).
There is also no weight based load balancing (the srv record with the
lowest prio and highest weight will always be selected).
The dns fall-back part references directly records in the dns cache from
tm (which saves a lot of memory) and hence it is disabled if the dns
cache is off.
Andrei