On ons, 2008-06-25 at 21:53 +0100, Bradley Kite wrote:
> Internal DNS Statistics:
>
> The Queue:
> DELAY SINCE
> ID SIZE SENDS FIRST SEND LAST SEND
> ------ ---- ----- ---------- ---------
> 0x2dd7 39 2597011 40711.226 0.002
> 0x10ec 40 2597905 42519.743 0.002
> 0x53d8 32 1 6584.452 6584.452
> 0xb10c 36 1 39209.722 39209.722
This looks very odd.. you have two entries which is retried a lot at a
high rate, and a lot of entries which got queried only one and then not
retransmitted even when their timer has been running for very long..
I tried reproducing this, but I only get proper retransmission of all
entries in the queue like expected..
Looking in the code.. a guess is that there is a TCP fallback on those
two, and that something go wrong during that. By reading the code it
looks like the DNS request then starts to bounces between UDP/TCP
without any bounds or delays..
But not sure why it stops the other DNS lookups dead so there seems to
be two issues here.
What is dns_retransmit_interval set to in your squid.conf?
And can you please file a bug on this.
http://www.squid-cache.org/bugs/
Regards
Henrik
This archive was generated by hypermail 2.2.0 : Thu Jun 26 2008 - 12:00:05 MDT