[Linux-HA] DNS problem with heartbeat
david.lang at digitalinsight.com
Thu Jul 14 18:09:02 MDT 2005
On Thu, 14 Jul 2005, Christof Amelunxen wrote:
> Hi David,
> David Lang sagte:
>>> Anyway, why dont you just configure one of your nodes as DNS master and
>>> the other one as slave and tell the clients to use them both as primary
>>> and secondary. This way you wouldnt have to deal with cluster failover
>> The problem with this is that many (most??) clients will wait for the
>> primary to time out before attempting a connection to the secondary. I've
>> seen this add 30 seconds to all DNS requests when setup this way.
> Thanks for the hint. On UNIX systems (just checked that with linux and
> solaris) there is no real noticeable delay in dns resolutions if the
> primary is down but windows clients may behave differently.
also note that there can be a huge difference between pointing at an IP
that's up, but not running DNS (quick rejection) and pointing at an IP
address that doesn't respond (slow timeout)
>> being able to failover the IP address that everything is pointing at so
>> that the primary DNS address is always up is a much nicer solution.
> Switching the ip address may have advantages, that is true. But I would
> recommend him to use the bind replication mechanism anyway, to avoid the
> need for shared storage, data replication etc. and not let heartbeat
> manage the startup and shutdown of bind but keep it running on both nodes.
> I guess his problems are related to this in a way.
you can do this, but then you have to watch your zone time-to-live
settings, if your primary box stays down too long the secondary will no
longer consider itself authoritative and will stop responding as well.
There are two ways of constructing a software design. One way is to make it so simple that there are obviously no deficiencies. And the other way is to make it so complicated that there are no obvious deficiencies.
-- C.A.R. Hoare
More information about the Linux-HA