Document the new unreachable cache behavior

Update the documentaion to include information about how the cache's
exponential backoff works, and how to clear the cache.

diff --git a/doc/arm/config-auth.inc.rst b/doc/arm/config-auth.inc.rst
index b054fb0904faffbe43cc15a29ef80110cb8a7e87..77d1c7ae99eb43d6aae4d30609f05e3f306a3f3b 100644 (file)
--- a/doc/arm/config-auth.inc.rst
+++ b/doc/arm/config-auth.inc.rst
@@ -87,6 +87,13 @@ The numbers in parentheses in the following text refer to the numbered items in
    from the primary (as described in section 2 a. above). If the zone file has
    changed, propagation is practically immediate.
 
+.. Note:: When a primary server is unreachable, :iscman:`named` initially caches
+   that information for 10 seconds. After that initial period, :iscman:`named`
+   may try that server again; if the server remains unresponsive,
+   :iscman:`named` caches its unreachable status for up to 640 seconds using an
+   exponential backoff. During that time :iscman:`named` does not try to contact
+   the affected server. The cache can be cleared using :option:`rndc flush`.
+
 The authoritative samples all use NOTIFY but identify the statements used, so
 that they can be removed if not required.