London Node - long DNS lookup times

Hi Pat,

i’m seeing quite large DNS lookup times for my tests from the London node.


its been like this for a while now, which DNS provider is it using?


I don’t have access to the agent to check but if you capture a tcpdump you can see where the DNS requests are going (and how long they are taking on the wire in case it’s not a network-level issue).

Hmm, a bit puzzled.

if i use wireshark to show the average dns.time (using the io graph) the average time is under 100ms.
The max time is shown as 300ms, this doesn’t match what WPT is telling me.

The long DNS times as shown in WPT have significantly increased the overall times from the London node. I did notice that there maybe a correlation with the resolution of akamai dns.

e.g is a CNAME for which is a CNAME for some * host.



Are you using wireshark locally or checking the “capture network packet trace (tcpdump)” box in the advanced settings and looking at the traces from the WPT agent?

The CPU graph on the agent looked like it was pegged so I was wondering if there was a test agent issue of some kind that was making them look longer than they really were on the wire.

Hi pat,

i opened the tcpdump file in wireshark.

then opened statistics → io graph
selected “Advanced…” as Y Axis unit value
then in graph 1:
filter on dns.time
calc Avg(*) dns.time

I repeated the above for Min and Max values on graph 2 and 3 respectively.

couldn’t see anything that suggested DNS resolution was slow.