[Pdns-users] Recursor doesn't wait long enough for the reply
miesi at pc-h.de
Thu Sep 1 14:09:01 UTC 2016
As a side note from my experience: Be very careful and don't set it high.
a) setting network-timeout to very high values like 20s will cause your
recursor to run out of sockets (Because they are all in use by
connections to slow or dead authoritative Servers)
b) Clients using your recursor have their own timeouts and don't care
for your recursors answer if the it is received after their timeout.
They will have retried, hitting another pdns-recursor thread consuming
another socket on the recursor.
So, don't set network-timeout higher than the time-out on your clients.
If you have problems with certain zones, just add
to your forward zones and that stops eating up sockets. Yes, google
delivers potential outdated data, but it does fast.
That is by far not a perfect solution but sometimes things must stop
Am 01.09.2016 um 12:34 schrieb Laszlo Toth:
> Replying for myself:
> I was looking for the network-timeout parameter.
> On 2016-09-01 11:17, Laszlo Toth wrote:
>> I have a problem with pdns-recursor. It looks like if the recursive
>> forwarder (that is set with the forward-zones-recurse parameter) is
>> taking longer than usual to respond the Recursor stops waiting for it.
>> My experiment showed this is 1 second. Is there a way to configure
>> this timeout? I have to use a couple of domains where the TTL is 30
>> (or 60 seconds) and their authoriative servers are very slow.
>> The error looks like this in tcpdump:
>> 13:56:18.082578 IP pdns.server.local.30044 > dns1.resolv.to.domain:
>> 39835+ A? example.com. (42)
>> 13:56:19.737475 IP dns1.resolv.to.domain > odns.server.local.30044:
>> 39835 1/0/0 A 126.96.36.199 (58)
>> 13:56:19.737511 IP pdns.server.local > dns1.resolv.to: ICMP
>> pdns.server.local udp port 30044 unreachable, length 94
>> Pdns-users mailing list
>> Pdns-users at mailman.powerdns.com
> Pdns-users mailing list
> Pdns-users at mailman.powerdns.com
More information about the Pdns-users