rndc flushname not working
lists+isc.org at elitists.org
Fri Apr 10 17:52:18 UTC 2015
On Fri, Apr 10, 2015 at 4:07 AM, Chris Buxton <clists at buxtonfamily.us> wrote:
> On Apr 9, 2015, at 4:25 PM, Frank Even <lists+isc.org at elitists.org> wrote:
>> Is there any place I can look to get a definitive answer in what cases
>> "flushname" will and will not work? I've been digging around in lists
>> and docs and can't seem to find any definitive answers. I've been
>> having odd troubles clearing a name from a cache and after even
>> clearing the name and the name that the name servers was attached to,
>> still had to flush the entire cache to get resolution working properly
>> on that domain again.
> 'rndc flushname' will clear a single name (all records, all types) out of the normal recursion cache. However, as noted by Mark last December, if an authoritative server for the domain is returning bad answers, the bad answer might come back on the next query for that rrset or that name.
> If you want to clear an entire domain out of cache (example.com and all of its child names), use 'rndc flushtree' instead.
flushtree does not work in the current distributed version of bind in
EL6 packages. In this particular case, when the issue came to me, the
name servers for the domain were able to return the result with no
problems, other caching servers throughout the company had no issues,
but this group of servers that apparently had tests run against the
domain prior to it being fully setup could not resolve it and nothing
short of a full flush of the cache would fix the issue. That leads me
to believe there is some data that gets held in cache not cleared by a
flushname that is relevant but not documented.
More information about the bind-users